<?xml version="1.0" encoding="UTF-8"?><rss xmlns:dc="http://purl.org/dc/elements/1.1/" xmlns:content="http://purl.org/rss/1.0/modules/content/" xmlns:atom="http://www.w3.org/2005/Atom" version="2.0" xmlns:cc="http://cyber.law.harvard.edu/rss/creativeCommonsRssModule.html">
    <channel>
        <title><![CDATA[Stories by Shalini Shree on Medium]]></title>
        <description><![CDATA[Stories by Shalini Shree on Medium]]></description>
        <link>https://medium.com/@shreeshalini2912?source=rss-239affa7bb58------2</link>
        <image>
            <url>https://cdn-images-1.medium.com/fit/c/150/150/0*2sCdwNNxUlxEbKWj</url>
            <title>Stories by Shalini Shree on Medium</title>
            <link>https://medium.com/@shreeshalini2912?source=rss-239affa7bb58------2</link>
        </image>
        <generator>Medium</generator>
        <lastBuildDate>Sun, 22 Nov 2026 18:39:22 GMT</lastBuildDate>
        <atom:link href="https://medium.com/@shreeshalini2912/feed" rel="self" type="application/rss+xml"/>
        <webMaster><![CDATA[yourfriends@medium.com]]></webMaster>
        <atom:link href="http://medium.superfeedr.com/" rel="hub"/>
        <item>
            <title><![CDATA[Causality Quiz Question]]></title>
            <link>https://medium.com/@shreeshalini2912/causality-quiz-question-1da2ad91869b?source=rss-239affa7bb58------2</link>
            <guid isPermaLink="false">https://medium.com/p/1da2ad91869b</guid>
            <category><![CDATA[causality]]></category>
            <category><![CDATA[causal-inference]]></category>
            <category><![CDATA[quiz]]></category>
            <dc:creator><![CDATA[Shalini Shree]]></dc:creator>
            <pubDate>Thu, 27 Apr 2023 03:08:24 GMT</pubDate>
            <atom:updated>2023-04-27T03:08:24.566Z</atom:updated>
            <content:encoded><![CDATA[<figure><img alt="" src="https://cdn-images-1.medium.com/max/768/1*cz-8WTC6v3QPYG7GVOoxaQ.jpeg" /></figure><p>1. Which of the following is NOT a necessary condition for establishing causation?</p><p>a) Temporal precedence</p><p>b) Strength of association</p><p>c) Correlation</p><p>d) Non-spuriousness</p><p><strong>Answer: c) Correlation</strong></p><p>2. In the context of causal inference, what does the term “counterfactual” refer to?</p><p>a) The observed outcome</p><p>b) The unobserved outcome without the treatment</p><p>c) The observed outcome with the treatment</p><p>d) The difference between treated and control groups</p><p><strong>Answer: b) The unobserved outcome without the treatment</strong></p><p>3. What is the main purpose of using matching methods in causal inference?</p><p>a) To balance the distribution of confounding variables across treatment and control groups</p><p>b) To estimate the causal effect using an exogenous instrument</p><p>c) To control for temporal trends</p><p>d) To exploit discontinuities in the assignment of treatment</p><p><strong>Answer: a) To balance the distribution of confounding variables across treatment and control groups</strong></p><p>4. A _________ variable is an external factor that affects both the treatment and outcome variables, leading to a spurious or biased relationship between them.</p><p><strong>Answer: confounding</strong></p><p>5. _________ score matching aims to create comparable treatment and control groups by matching individuals based on their propensity scores, which represent the estimated probability of receiving treatment given observed characteristics.</p><p><strong>Answer: Propensity</strong></p><p>6. What is a confounding variable?</p><p><strong>Answer: A confounding variable is an external factor that affects both the treatment and outcome variables, leading to a spurious or biased relationship between them.</strong></p><p>7. What is the key difference between correlation and causation?</p><p><strong>Answer: Correlation refers to the statistical relationship between two variables, while causation implies that a change in one variable directly causes a change in the other variable.</strong></p><p>8. True or False: The average treatment effect (ATE) is the average difference in outcomes between treatment and control groups specifically for those who received the treatment.</p><p><strong>Answer: False (It is the average treatment effect on the treated (ATT) that represents this difference.)</strong></p><p>9. True or False: Randomization in a controlled experiment helps to balance the distribution of confounding variables across treatment and control groups.</p><p><strong>Answer: True</strong></p><p>10. In causal inference, the primary goal is to determine the _________ effect of a treatment or intervention on an outcome by comparing the outcomes with and without the treatment.</p><p><strong>Answer: causal</strong></p><h3>REFERENCES</h3><ul><li><a href="https://github.com/shoepaladin/causalinference_crashcourse/blob/main/Notebooks/1%20Foundations.ipynb">https://github.com/shoepaladin/causalinference_crashcourse/blob/main/Notebooks/1%20Foundations.ipynb</a></li><li><a href="https://github.com/shardulchavan/Crash_Course_in_Causality/blob/main/Crash_Course_in_Causality.ipynb">https://github.com/shardulchavan/Crash_Course_in_Causality/blob/main/Crash_Course_in_Causality.ipynb</a></li><li><a href="https://www.bradyneal.com/causal-inference-course">https://www.bradyneal.com/causal-inference-course</a></li></ul><h3>MIT License</h3><p>All code in this notebook is available as open source through the MIT license.</p><p>All text and images are free to use under the Creative Commons Attribution 3.0 license. <a href="https://creativecommons.org/licenses/by/3.0/us/">https://creativecommons.org/licenses/by/3.0/us/</a></p><p>These licenses let people distribute, remix, tweak, and build upon the work, even commercially, as long as they give credit for the original creation.</p><p>Copyright 2023 AI Skunks <a href="https://github.com/aiskunks">https://github.com/aiskunks</a></p><p>Permission is hereby granted, free of charge, to any person obtaining a copy of this software and associated documentation files (the “Software”), to deal in the Software without restriction, including without limitation the rights to use, copy, modify, merge, publish, distribute, sublicense, and/or sell copies of the Software, and to permit persons to whom the Software is furnished to do so, subject to the following conditions:</p><p>The above copyright notice and this permission notice shall be included in all copies or substantial portions of the Software.</p><p>THE SOFTWARE IS PROVIDED “AS IS”, WITHOUT WARRANTY OF ANY KIND, EXPRESS OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL THE AUTHORS OR COPYRIGHT HOLDERS BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE.</p><img src="https://medium.com/_/stat?event=post.clientViewed&referrerSource=full_rss&postId=1da2ad91869b" width="1" height="1" alt="">]]></content:encoded>
        </item>
        <item>
            <title><![CDATA[Crash Course in Causality Worked Examples]]></title>
            <link>https://medium.com/@shreeshalini2912/crash-course-in-causality-worked-examples-a8382ebbf8d6?source=rss-239affa7bb58------2</link>
            <guid isPermaLink="false">https://medium.com/p/a8382ebbf8d6</guid>
            <category><![CDATA[causation]]></category>
            <category><![CDATA[causal-data-science]]></category>
            <category><![CDATA[causality]]></category>
            <category><![CDATA[causal-inference]]></category>
            <dc:creator><![CDATA[Shalini Shree]]></dc:creator>
            <pubDate>Wed, 26 Apr 2023 22:49:52 GMT</pubDate>
            <atom:updated>2023-04-26T22:50:32.853Z</atom:updated>
            <content:encoded><![CDATA[<p><em>Shalini Shree, Nik Brown</em></p><p>Abstract: Causality is the study of how one variable influences another variable, with a focus on identifying cause-and-effect relationships. It involves understanding the underlying mechanisms and factors that lead to observed relationships between variables and distinguishing between spurious correlations and genuine causal relationships. Causality is essential in many fields, including social sciences, economics, public health, and medicine, as it allows us to make informed decisions and take action to address the root causes of problems. Causal inference techniques are used to establish causal relationships, taking into account potential confounding factors and other sources of bias.</p><h3>We will be discussing the application of DoWhy library used to find causal inference on two different datasets</h3><p>DoWhy is a Python library for causal inference that provides a unified interface for estimating causal effects using different methods. It is designed to facilitate causal analysis and automate the steps involved in causal inference, including the identification of causal relationships, the estimation of causal effects, and the testing of assumptions.</p><p>DoWhy allows us to easily specify causal models using a high-level language that is based on the graphical model notation. It then automatically identifies the causal effect to be estimated and generates a corresponding statistical model. It supports a wide range of methods for causal inference, including regression, matching, weighting, and instrumental variables.</p><h4>DATASET — 1</h4><p>The dataset contains information about the mortality rates in France between the years 2010 and 2015. It includes data about the number of deaths for each age group and sex, as well as the underlying cause of death. The dataset also provides information on the level of cause, which indicates the specificity of the cause of death.</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*34rsAwpbRrDk93XyU3PJdw.png" /></figure><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*6R55PLv_62W6IcnxDwLtiQ.png" /></figure><p>This output is the result of a causal inference analysis using the backdoor propensity score matching method to estimate the causal effect of the concept of smoking (concept_id_french) on the total number of deaths in France (total_yr_deaths_FRANCE).</p><p>The identified estimand indicates that the analysis is estimating the non-parametric average treatment effect (ATE), which is the difference in the mean outcome between the treated and untreated groups in the population.</p><p>The realized estimand shows the model specification used to estimate the causal effect, which includes the covariates concept_id_french and level_of_cause in a linear regression model (total_yr_deaths_FRANCE~concept_id_french+level_of_cause).</p><p>The target units for the estimate are the average treatment effect (ATE), which represents the average causal effect of smoking on the total number of deaths in France across all individuals in the population.</p><p>The estimate value is the mean of the estimated causal effect, which in this case is 11379.99999999997. This means that smoking is estimated to cause an average increase of approximately 11,380 deaths per year in France.</p><h4>DATASET -2</h4><p>The Titanic dataset is a classic dataset in the field of machine learning and data science. It contains information about the passengers onboard the Titanic ship, which famously sank in 1912. The dataset includes features such as age, sex, passenger class, and ticket fare, as well as whether each passenger survived or not. The main objective of the dataset is to predict whether a passenger survived or not based on the given features. The dataset contains 891 observations or rows and 12 variables or columns. It is a popular dataset for practicing data wrangling, exploratory data analysis, and predictive modeling techniques.</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*gsDAjMyAZnyeHJeDxJJJbg.png" /></figure><p>The output of this code shows us the impact of age on survival, while considering other variables that may have an effect on survival such as Pclass and Sex. This can help us determine if age has a direct effect on survival on the Titanic, after accounting for the influence of other variables that may be related to both age and survival.</p><p>The estimate value shows the estimated ATE, which is the average difference in survival probabilities between the treated group and the control group over the entire population. In this case, the ATE is estimated to be -0.005, which means that age has a very small negative effect on survival probability, holding sex and passenger class constant.</p><h3><strong>REFERENCES</strong></h3><ul><li><a href="https://github.com/shoepaladin/causalinference_crashcourse/blob/main/Notebooks/1%20Foundations.ipynb">https://github.com/shoepaladin/causalinference_crashcourse/blob/main/Notebooks/1%20Foundations.ipynb</a></li><li><a href="https://github.com/shardulchavan/Crash_Course_in_Causality/blob/main/Crash_Course_in_Causality.ipynb">https://github.com/shardulchavan/Crash_Course_in_Causality/blob/main/Crash_Course_in_Causality.ipynb</a></li><li><a href="https://www.bradyneal.com/causal-inference-course">https://www.bradyneal.com/causal-inference-course</a></li></ul><h3>MIT License</h3><p>All code in this notebook is available as open source through the MIT license.</p><p>All text and images are free to use under the Creative Commons Attribution 3.0 license. <a href="https://creativecommons.org/licenses/by/3.0/us/">https://creativecommons.org/licenses/by/3.0/us/</a></p><p>These licenses let people distribute, remix, tweak, and build upon the work, even commercially, as long as they give credit for the original creation.</p><p>Copyright 2023 AI Skunks <a href="https://github.com/aiskunks">https://github.com/aiskunks</a></p><p>Permission is hereby granted, free of charge, to any person obtaining a copy of this software and associated documentation files (the “Software”), to deal in the Software without restriction, including without limitation the rights to use, copy, modify, merge, publish, distribute, sublicense, and/or sell copies of the Software, and to permit persons to whom the Software is furnished to do so, subject to the following conditions:</p><p>The above copyright notice and this permission notice shall be included in all copies or substantial portions of the Software.</p><p>THE SOFTWARE IS PROVIDED “AS IS”, WITHOUT WARRANTY OF ANY KIND, EXPRESS OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL THE AUTHORS OR COPYRIGHT HOLDERS BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE.</p><img src="https://medium.com/_/stat?event=post.clientViewed&referrerSource=full_rss&postId=a8382ebbf8d6" width="1" height="1" alt="">]]></content:encoded>
        </item>
        <item>
            <title><![CDATA[Crash Course in Causality: Smoking Leads to Death?]]></title>
            <link>https://medium.com/aiskunks/crash-course-in-causality-smoking-leads-to-death-4bedea49be43?source=rss-239affa7bb58------2</link>
            <guid isPermaLink="false">https://medium.com/p/4bedea49be43</guid>
            <category><![CDATA[causality]]></category>
            <category><![CDATA[causal-inference]]></category>
            <category><![CDATA[causal-data-science]]></category>
            <category><![CDATA[data-science]]></category>
            <dc:creator><![CDATA[Shalini Shree]]></dc:creator>
            <pubDate>Sun, 23 Apr 2023 02:55:58 GMT</pubDate>
            <atom:updated>2023-04-28T20:55:31.130Z</atom:updated>
            <content:encoded><![CDATA[<p><em>Shalini Shree, Nik Brown</em></p><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*d_DHNFHuKoJ9O5NHLTACwQ.jpeg" /></figure><p>In today’s data-driven world, the ability to extract meaningful insights from complex datasets has become a vital skill. Causal inference, a cornerstone of modern data science, is a powerful tool that empowers researchers and decision-makers to understand the underlying relationships between variables and determine the impact of interventions. At its core, causal inference helps us answer the fundamental question: “What would happen if we did X?” This intriguing branch of statistics transcends mere correlation, enabling us to establish cause-and-effect relationships that are paramount for informed decision-making in fields ranging from healthcare and economics to public policy and marketing.</p><p>This article will utilize a sample dataset that includes data on deaths caused by smoking among different age groups and genders in France. The objective is to explore whether there is a causal relationship between smoking and death or if there is only a correlation between the two variables.</p><h3>What is causality?</h3><p>Causality refers to the relationship between <strong>cause</strong> and <strong>effect</strong>. In other words, it refers to the idea that one event (the cause) can directly or indirectly result in another event (the effect). Causal relationships are important in many fields, including science, economics, and social sciences.</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/363/1*nG6Vo8y5yvb17HfIZRVoXg.png" /></figure><p>In causality, we use the <strong>treatment</strong> and <strong>outcome </strong>variables to investigate whether a causal relationship exists between them. The treatment variable is the variable we want to investigate as a potential cause of the outcome variable. The outcome variable is the variable we want to investigate as the effect of the treatment variable.</p><p>We start by defining a hypothesis about the causal relationship between the treatment and outcome variables. We then collect data on both variables and any potential confounding variables. We then use statistical methods, such as regression analysis or experimental design, to estimate the causal effect of the treatment variable on the outcome variable, while controlling for the potential confounding variables. The goal is to determine whether the treatment variable has a significant effect on the outcome variable, and if so, the size and direction of the effect. If a significant causal relationship is found between the treatment and outcome variables, this can provide evidence for a potential cause-and-effect relationship between them.</p><blockquote>A confounding variable is a variable that affects both the independent variable (treatment) and the dependent variable (outcome), making it difficult to determine the true causal effect between them by distorting the relationship between treatment and outcome, leading to incorrect conclusions about causality.</blockquote><h3>Why is causal inference important?</h3><p>Causal inference is important because it allows us to understand the relationship between variables and determine whether one variable is actually causing changes in another. This information is crucial in many fields, including medicine, public health, economics, and social sciences, as it can help us make informed decisions and create effective interventions.</p><p>Furthermore, causal inference is important because it allows us to move beyond simply observing correlations between variables and making statements about the actual causal relationship between them. This helps us avoid making incorrect assumptions about the relationship between variables, which can lead to ineffective interventions and wasted resources.</p><blockquote>For example, in medicine, causal inference can help determine whether a particular treatment is effective in improving patient outcomes, or whether a certain risk factor is contributing to the development of a disease. In public health, causal inference can help identify the factors that are contributing to the spread of a disease and guide the development of interventions to control its spread.</blockquote><h3>What are the differences and similarities between causality and correlation?</h3><figure><img alt="" src="https://cdn-images-1.medium.com/max/457/1*v3bNyT5sjP3pQC8SCIJgrg.png" /></figure><p>Causality and correlation are both concepts that describe the relationship between variables, but they differ in important ways.</p><p>Correlation is a statistical measure that describes the strength and direction of the relationship between two variables. It measures how much one variable changes as the other variable changes, and ranges from -1 to +1, where -1 indicates a strong negative correlation, 0 indicates no correlation, and +1 indicates a strong positive correlation. Correlation does not imply causation, meaning that just because two variables are correlated, it does not necessarily mean that changes in one variable cause changes in the other.</p><p>Causality, on the other hand, refers to the relationship between cause and effect. It implies that changes in one variable cause changes in another variable. Causality can be inferred through various methods, including randomized experiments, natural experiments, and observational studies using causal inference techniques. Establishing causality requires more evidence than establishing correlation, as it involves ruling out alternative explanations for the observed relationship.</p><h3>Let&#39;s take a look at our Example dataset</h3><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*JvjEWeU48TuqxnqomxR3xA.png" /></figure><figure><img alt="" src="https://cdn-images-1.medium.com/max/854/1*bBfIp9G_RkMBO5r3z_RMCg.png" /></figure><figure><img alt="" src="https://cdn-images-1.medium.com/max/731/1*aaY63-Rmj-8s8_cHPSY7zw.png" /></figure><figure><img alt="" src="https://cdn-images-1.medium.com/max/744/1*3sRuQOc-ec5ZlILw8l8tmA.png" /></figure><figure><img alt="" src="https://cdn-images-1.medium.com/max/515/1*4liwprI0VukpQ902-Jxclw.png" /></figure><figure><img alt="" src="https://cdn-images-1.medium.com/max/977/1*0dlj4kG-CCvhdG4UaL0A7g.png" /></figure><h3>What is DoWhy Library?</h3><p>DoWhy is a Python library for causal inference that provides a unified interface for estimating causal effects using different methods. It is designed to facilitate causal analysis and automate the steps involved in causal inference, including the identification of causal relationships, the estimation of causal effects, and the testing of assumptions.</p><p>DoWhy allows us to easily specify causal models using a high-level language that is based on the graphical model notation. It then automatically identifies the causal effect to be estimated and generates a corresponding statistical model. It supports a wide range of methods for causal inference, including regression, matching, weighting, and instrumental variables.</p><p>The main advantage of using DoWhy is that it allows researchers and data scientists to perform causal inference without having to manually specify and estimate causal models, which can be a time-consuming and error-prone process. By automating many of the steps involved in causal inference, DoWhy can help to make causal analysis more accessible to a wider range of users and enable more accurate and reliable causal inference.</p><p><em>Let us use the DoWhy library to calculate the causal effect for our dataset.</em></p><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*34rsAwpbRrDk93XyU3PJdw.png" /></figure><p>We are creating a causal model using the DoWhy library. The dataset smoking_francedata_copy is being used, concept_id_french as the treatment variable and total_yr_deaths_FRANCE as the outcome variable. The variable level_of_cause is included as a common cause of the treatment and outcome variables. No instrument variable is included in this model. The resulting model object can be used to estimate the causal effect of the treatment on the outcome using various methods provided by the DoWhy library.</p><p>We are then identifying the causal effect using the identify_effect method of the CausalModel class from the DoWhy library. The proceed_when_unidentifiable=True argument allows the method to proceed even when a causal effect cannot be identified, which is useful for exploring the data. Once the causal effect has been identified, we estimate it using the estimate_effect method with the identified estimand and the backdoor.linear_regression method for causal inference. This method estimates the causal effect using linear regression with the backdoor adjustment technique.</p><blockquote>The backdoor criterion states that, in order to identify the causal effect of a treatment on an outcome, all backdoor paths between the treatment and the outcome must be blocked. A backdoor path is a path from the treatment to the outcome that contains at least one confounding variable.</blockquote><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*6R55PLv_62W6IcnxDwLtiQ.png" /></figure><p>This output is the result of a causal inference analysis using the backdoor propensity score matching method to estimate the causal effect of the concept of smoking (concept_id_french) on the total number of deaths in France (total_yr_deaths_FRANCE).</p><p>The identified estimand indicates that the analysis is estimating the non-parametric average treatment effect (ATE), which is the difference in the mean outcome between the treated and untreated groups in the population.</p><p>The realized estimand shows the model specification used to estimate the causal effect, which includes the covariates concept_id_french and level_of_cause in a linear regression model (total_yr_deaths_FRANCE~concept_id_french+level_of_cause).</p><p>The target units for the estimate are the average treatment effect (ATE), which represents the average causal effect of smoking on the total number of deaths in France across all individuals in the population.</p><p>The estimate value is the mean of the estimated causal effect, which in this case is 11379.99999999997. This means that smoking is estimated to cause an average increase of approximately 11,380 deaths per year in France.</p><h3>REFERENCES</h3><ul><li><a href="https://github.com/shoepaladin/causalinference_crashcourse/blob/main/Notebooks/1%20Foundations.ipynb">https://github.com/shoepaladin/causalinference_crashcourse/blob/main/Notebooks/1%20Foundations.ipynb</a></li><li><a href="https://github.com/shardulchavan/Crash_Course_in_Causality/blob/main/Crash_Course_in_Causality.ipynb">https://github.com/shardulchavan/Crash_Course_in_Causality/blob/main/Crash_Course_in_Causality.ipynb</a></li><li><a href="https://www.bradyneal.com/causal-inference-course">https://www.bradyneal.com/causal-inference-course</a></li></ul><h3>MIT License</h3><p>All code in this notebook is available as open source through the MIT license.</p><p>All text and images are free to use under the Creative Commons Attribution 3.0 license. <a href="https://creativecommons.org/licenses/by/3.0/us/">https://creativecommons.org/licenses/by/3.0/us/</a></p><p>These licenses let people distribute, remix, tweak, and build upon the work, even commercially, as long as they give credit for the original creation.</p><p>Copyright 2023 AI Skunks <a href="https://github.com/aiskunks">https://github.com/aiskunks</a></p><p>Permission is hereby granted, free of charge, to any person obtaining a copy of this software and associated documentation files (the “Software”), to deal in the Software without restriction, including without limitation the rights to use, copy, modify, merge, publish, distribute, sublicense, and/or sell copies of the Software, and to permit persons to whom the Software is furnished to do so, subject to the following conditions:</p><p>The above copyright notice and this permission notice shall be included in all copies or substantial portions of the Software.</p><p>THE SOFTWARE IS PROVIDED “AS IS”, WITHOUT WARRANTY OF ANY KIND, EXPRESS OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL THE AUTHORS OR COPYRIGHT HOLDERS BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE</p><img src="https://medium.com/_/stat?event=post.clientViewed&referrerSource=full_rss&postId=4bedea49be43" width="1" height="1" alt=""><hr><p><a href="https://medium.com/aiskunks/crash-course-in-causality-smoking-leads-to-death-4bedea49be43">Crash Course in Causality: Smoking Leads to Death?</a> was originally published in <a href="https://medium.com/aiskunks">AI Skunks</a> on Medium, where people are continuing the conversation by highlighting and responding to this story.</p>]]></content:encoded>
        </item>
        <item>
            <title><![CDATA[Crash Course in Forecasting Worked Examples]]></title>
            <link>https://medium.com/@shreeshalini2912/crash-course-in-forecasting-worked-examples-2d4f301723eb?source=rss-239affa7bb58------2</link>
            <guid isPermaLink="false">https://medium.com/p/2d4f301723eb</guid>
            <category><![CDATA[time-series-analysis]]></category>
            <category><![CDATA[time-series-forecasting]]></category>
            <category><![CDATA[data-science]]></category>
            <category><![CDATA[machine-learning]]></category>
            <category><![CDATA[example]]></category>
            <dc:creator><![CDATA[Shalini Shree]]></dc:creator>
            <pubDate>Mon, 10 Apr 2023 03:30:58 GMT</pubDate>
            <atom:updated>2023-04-10T04:04:53.704Z</atom:updated>
            <content:encoded><![CDATA[<p><em>Shalini Shree, Nik Brown</em></p><p>Abstract: Time series analysis is a critical tool for understanding patterns and trends in data collected over time, with applications in various domains. This article presents the three fundamental steps involved in time series analysis: visualizing the data, detecting, and imputing missing values, and splitting the data into training and testing sets. Visualization helps identify patterns while missing data can lead to biased results, so imputation techniques are used. Splitting the data into training and testing sets helps ensure the model generalizes well on new data and avoids overfitting. These steps form the foundation for developing accurate models to forecast future trends and patterns.</p><h3>We will be discussing the application of GreyKite on 2 time-series data sets</h3><p>For data scientists, LinkedIn has released the time-series forecasting library Greykite. This library uses Silverkite as its main forecasting algorithm, which automates forecasting. GreyKite was created by LinkedIn to help its employees make decisions that are effective and based on time-series forecasting models. As the library also aids in output interpretation, it can end up becoming a go-to tool for most of the time-series forecasting. A Fairness Toolkit for explainability in machine learning was also made available by LinkedIn last year.</p><p>Throughout time, LinkedIn has used the Greykite library to build up the necessary infrastructure to manage peak traffic, establish business goals, and improve spending choices.</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/0*8HW73dI9wIPQYqgS.png" /><figcaption>GreyKite’s architecture is based on the Silverkite algorithm made by LinkedIn.</figcaption></figure><h3>DATASET — 1</h3><p>Smart Water Analytics: Time series forecasting using statistical models</p><p>The necessity to predict the water level in a waterbody (water spring, lake, river, or aquifer) to manage daily demand presents a challenge to water distribution firms. Water bodies are replenished in the fall and winter, but they begin to drain in the spring and summer. It is crucial to forecast the most effective water availability, in terms of level and water flow, for each day of the year in order to support the maintenance of the health of these water bodies.</p><p>Data on 4 different categories of water bodies are provided by the Acea Group.</p><p>Water Spring, Lake, River, Aquifer</p><p>Although the main goal is the same — to forecast the availability of water — the truth is that each water body has so many distinctive qualities that their properties are unrelated to one another.</p><p>We have used various methods like ARIMA, Prophet, GreyKite, and Neural Networks to forecast water levels in one of the aquifers (aquifer in Petrigrano).</p><p>Code for setting up the GreyKite model for Monthly Frequency</p><pre># Specifies dataset information<br>metadata = MetadataParam(<br>     time_col=&quot;ts&quot;,  # name of the time column<br>     value_col=&quot;y&quot;,  # name of the value column<br>     freq=&quot;D&quot;  #&quot;MS&quot; for Montly at start date, &quot;H&quot; for hourly, &quot;D&quot; for daily, &quot;W&quot; for weekly, etc.<br> )<br>forecaster = Forecaster()<br>result = forecaster.run_forecast_config(<br>     df=df,<br>     config=ForecastConfig(<br>         model_template=ModelTemplateEnum.SILVERKITE.name,<br>         forecast_horizon=50,  # forecasts 100 steps ahead<br>         coverage=0.95,  # 95% prediction intervals<br>         metadata_param=metadata<br>    )<br>)</pre><pre>ts = result.timeseries<br>fig = ts.plot()<br>plotly.io.show(fig)</pre><p>The x-axis of the plot shows the time points for the time series, while the y-axis shows the depth_to_groundwater value</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/718/0*cpuExC3Lgn3dSu8B.png" /></figure><p>Plot for the target variable (Produced Units of Electricity index (y)) vs Time (Yearly)</p><p>The plot below shows the upper and lower prediction intervals for the forecasted values which is depth_to_groundwater where the blue line predicts the forecasted value and the red line is the actual value. This graph shows a minimal difference between the actual and predicted value of test data while evaluating the performance of the Silverkite algorithm in forecasting the time-series data.</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/696/0*VvZb-zfwoYmG0Ig9.png" /></figure><pre>backtest = result.backtest<br>fig = backtest.plot()<br>plotly.io.show(fig)</pre><pre>backtest_eval = defaultdict(list)<br>for metric, value in backtest.train_evaluation.items():<br>    backtest_eval[metric].append(value)<br>    backtest_eval[metric].append(backtest.test_evaluation[metric])<br>metrics = pd.DataFrame(backtest_eval, index=[&quot;train&quot;, &quot;test&quot;]).T<br>metrics</pre><figure><img alt="" src="https://cdn-images-1.medium.com/max/782/0*orcM1ujqeZAit3xw.png" /></figure><p>The plot below shows the upper and lower prediction intervals for the forecasted values which is depth_to_groundwater where the blue line predicts the forecasted value and the red line is the actual value. This graph shows a minimal difference between the actual and predicted value of validation data while evaluating the performance of the Silverkite algorithm in forecasting the time series data.</p><pre>forecast = result.forecast<br>fig = forecast.plot()<br>plotly.io.show(fig)</pre><p>The plot below shows the upper and lower prediction intervals for the forecasted values which is depth_to_groundwater where the blue line predicts the forecasted value and the red line is the actual value. This graph shows a minimal difference between the actual and predicted value of validation data while evaluating the performance of the Silverkite algorithm in forecasting the time series data.</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/0*R8DjknQgd2kMRzDf.png" /></figure><p>The graph below shows different time components such as trends at different intervals like weekly, quarterly, and yearly. The weekly trend indicates a peak in water level at the beginning, while the quarterly trend shows a peak towards the end of the quarter. However, in the yearly plot, the trend reaches its highest point in the middle of the year. Additionally, the model also shows the events and residuals of groundwater depth over time, highlighting the variations in the target variable features.</p><pre>fig = forecast.plot_components()<br>plotly.io.show(fig)</pre><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/0*Zr-ddqVO3UrEstN-.png" /></figure><h3>DATASET — 2</h3><p>TIME SERIES DATA FOR INDEX OF ELECTRIC AND GAS UTILITIES PRODUCTION</p><p>Industrial production of electric and gas utilities in the United States, from the years 1985–2020, with our frequency being Monthly production output.</p><p>Frequency: Monthly</p><p>The industrial production (IP) index measures the real output of all relevant establishments located in the United States, regardless of their ownership, but not those located in U.S. territories.</p><p>The Federal Reserve’s monthly index of industrial production and the related capacity indexes and capacity utilization rates cover manufacturing, mining, and electric and gas utilities. The industrial sector, together with construction, accounts for the bulk of the variation in national output over the course of the business cycle. The industrial detail provided by these measures helps illuminate structural developments in the economy.</p><p>The production index measures real output and is expressed as a percentage of real output in a base year, currently 2017. The capacity index, which is an estimate of sustainable potential output, is also expressed as a percentage of actual output in 2017. The production indexes are computed as Fisher indexes since 1972; the weights are based on annual estimates of value added. The rate of capacity utilization equals the seasonally adjusted output index expressed as a percentage of the related capacity index.</p><p>SOURCE: <a href="https://fred.stlouisfed.org/series/IPG2211A2N">https://fred.stlouisfed.org/series/IPG2211A2N</a></p><pre># Specifies dataset information<br>metadata = MetadataParam(<br>     time_col=&quot;ts&quot;,  # name of the time column<br>     value_col=&quot;y&quot;,  # name of the value column<br>     freq=&quot;MS&quot;  #&quot;MS&quot; for Montly at start date, &quot;H&quot; for hourly, &quot;D&quot; for daily, &quot;W&quot; for weekly, etc.<br> )<br>forecaster = Forecaster()<br>result = forecaster.run_forecast_config(<br>     df=df,<br>     config=ForecastConfig(<br>         model_template=ModelTemplateEnum.SILVERKITE.name,<br>         forecast_horizon=50,  # forecasts 50 steps ahead<br>         coverage=0.95,  # 95% prediction intervals<br>         metadata_param=metadata<br>    )<br>)</pre><pre>ts = result.timeseries<br>fig = ts.plot()<br>plotly.io.show(fig)</pre><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/0*8mZDoDeQIQdCzUHp.png" /><figcaption>Plot for the target variable (Produced Units of Electricity index (y)) vs Time (Yearly)</figcaption></figure><pre>backtest = result.backtest<br>fig = backtest.plot()<br>plotly.io.show(fig)</pre><pre>backtest_eval = defaultdict(list)<br>for metric, value in backtest.train_evaluation.items():<br>    backtest_eval[metric].append(value)<br>    backtest_eval[metric].append(backtest.test_evaluation[metric])<br>metrics = pd.DataFrame(backtest_eval, index=[&quot;train&quot;, &quot;test&quot;]).T<br>metrics</pre><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/0*bIq-QwnDsy86b3Is.png" /><figcaption>Plot for forecasted vs actual values On test Data</figcaption></figure><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/0*ZsrCH4xXQ3MkIw5u.png" /><figcaption>Output Metrics on Held Out Dataset</figcaption></figure><pre>forecast = result.forecast<br>fig = forecast.plot()<br>plotly.io.show(fig)</pre><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/0*KdfGFAbowYMFYhaF.png" /><figcaption>Plot for forecasted vs actual values on the validation set</figcaption></figure><p>GreyKite Model clearly depicts components of time (trend, events, and residual) w.r.t time where these features of the target variable vary. The trend change point indicates that in 2000 the trend started to flatten out.</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/936/0*qIDsn5VbTz3ZAA_w.png" /><figcaption>Trends, Events, and Residuals for the 2nd Data Set.</figcaption></figure><h3>Conclusion</h3><p>We have gone over GreyKite Libraries implementation over 2 worked examples for different datasets.</p><h3>MIT License</h3><p>All code in this notebook is available as open source through the MIT license.</p><p>All text and images are free to use under the Creative Commons Attribution 3.0 license. <a href="https://creativecommons.org/licenses/by/3.0/us/">https://creativecommons.org/licenses/by/3.0/us/</a></p><p>These licenses let people distribute, remix, tweak, and build upon the work, even commercially, as long as they give credit for the original creation.</p><p>Copyright 2023 AI Skunks <a href="https://github.com/aiskunks">https://github.com/aiskunks</a></p><p>Permission is hereby granted, free of charge, to any person obtaining a copy of this software and associated documentation files (the “Software”), to deal in the Software without restriction, including without limitation the rights to use, copy, modify, merge, publish, distribute, sublicense, and/or sell copies of the Software, and to permit persons to whom the Software is furnished to do so, subject to the following conditions:</p><p>The above copyright notice and this permission notice shall be included in all copies or substantial portions of the Software.</p><p>THE SOFTWARE IS PROVIDED “AS IS”, WITHOUT WARRANTY OF ANY KIND, EXPRESS OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL THE AUTHORS OR COPYRIGHT HOLDERS BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE.</p><h3>References</h3><p><a href="https://www.kaggle.com/code/sandhyakrishnan02/time-series-analysis-and-forecasting">Time Series Forecasting Kaggle</a></p><p><a href="https://colab.research.google.com/drive/1FyhK0hhscSQqHfo227QvarzcWJAlIDDy?usp=sharing#scrollTo=pxSFVtPFdIhC">Choosing order with AIC and BIC</a></p><p><a href="https://colab.research.google.com/drive/19QlC6oNDTK5VMxvufI1uETQ7AXohlLDt?usp=sharing#scrollTo=6nZDXq48ZzTa">Data Visualization for Time Series</a></p><p><a href="https://medium.com/geekculture/greykite-the-new-forecasting-library-from-linkedin-case-bitcoin-price-prediction-981d21e79b31">Reference for GreyKite</a></p><img src="https://medium.com/_/stat?event=post.clientViewed&referrerSource=full_rss&postId=2d4f301723eb" width="1" height="1" alt="">]]></content:encoded>
        </item>
        <item>
            <title><![CDATA[Forecasting Quiz Question]]></title>
            <link>https://medium.com/@shreeshalini2912/forecasting-quiz-question-596eb7840cf9?source=rss-239affa7bb58------2</link>
            <guid isPermaLink="false">https://medium.com/p/596eb7840cf9</guid>
            <category><![CDATA[time-series-analysis]]></category>
            <category><![CDATA[machine-learning]]></category>
            <category><![CDATA[stationarity]]></category>
            <category><![CDATA[data-science]]></category>
            <category><![CDATA[arima]]></category>
            <dc:creator><![CDATA[Shalini Shree]]></dc:creator>
            <pubDate>Mon, 10 Apr 2023 03:05:05 GMT</pubDate>
            <atom:updated>2023-04-10T03:09:37.876Z</atom:updated>
            <content:encoded><![CDATA[<p>1. Which of the following is not a component of time series data?</p><p>A. Trend</p><p>B. Seasonality</p><p>C. Cyclicality</p><p>D. Outliers</p><p>Answer: A</p><p>2. What is the purpose of the autocorrelation function (ACF) in time series analysis?</p><p>A. To identify the presence of seasonality</p><p>B. To detect and remove outliers</p><p>C. To identify the order of the autoregressive (AR) model</p><p>D. To identify the order of the moving average (MA) model</p><p>Answer: C</p><p>3. The Akaike Information Criterion (AIC) is used to select the best ARIMA model for a given time series.</p><p>A. True</p><p>B. False</p><p>Answer: True</p><p>4. The seasonal decomposition of time series (STL) method decomposes a time series into ___________ components.</p><p>Answer: trend, seasonal, and error</p><p>5. Consider time series data that records the daily stock prices of a company. The data is shown in the table below:</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/1006/1*rv5Z94uVznCtZrARc5Q0xg.png" /></figure><p>a) Based on the above data, is the stock price series stationary or non-stationary?</p><p>b) If the series is non-stationary, what can you do to make it stationary?</p><p>c) What type of time series model can you use to forecast future stock prices?</p><p>Answer:</p><p>a) the stock price series has an upward trend, indicating that it is non-stationary.</p><p>b) To make the series stationary, we can apply differencing, which means taking the difference between consecutive values in the series. Here is the different series:</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/954/1*4qjwNS-VJu32OYQP_CJdtQ.png" /></figure><p>Now, we can see that the differenced series has no trend and appears to be stationary.</p><p>c) We can use an autoregressive integrated moving average (ARIMA) model to forecast future stock prices.</p><p>6. The autocorrelation function (ACF) measures the ___________ between a time series and its lagged values.</p><p>Answer: correlation</p><p>7. The difference between an AR model and an MA model is that the former uses past observations of the dependent variable as predictors, while the latter uses past errors as predictors.</p><p>A. True</p><p>B. False</p><p>Answer: True</p><p>8. What is the purpose of differencing in time series analysis?</p><p>A. To remove seasonality from the data</p><p>B. To transform the data into a stationary series</p><p>C. To detect outliers in the data</p><p>D. To identify the order of the ARIMA model</p><p>Answer: B</p><p>9. What is the difference between a stationary and non-stationary time series?</p><p>Answer: Stationarity is a desirable property of time series data because it simplifies the modeling and forecasting process. A stationary time series is one in which the statistical properties such as mean, variance, and covariance remain constant over time. On the other hand, a non-stationary time series is one in which the statistical properties change over time.</p><p>10. Which of the following is a limitation of time series models?</p><p>A. They are not suitable for short-term forecasting</p><p>B. They require a large amount of historical data to be effective</p><p>C. They assume that past patterns will continue into the future</p><p>D. They cannot handle non-linear relationships between variables</p><p>Answer: C</p><h3>References</h3><p><a href="https://www.kaggle.com/code/sandhyakrishnan02/time-series-analysis-and-forecasting">Time Series Forecasting Kaggle</a></p><p><a href="https://colab.research.google.com/drive/1FyhK0hhscSQqHfo227QvarzcWJAlIDDy?usp=sharing#scrollTo=pxSFVtPFdIhC">Choosing order with AIC and BIC</a></p><p><a href="https://colab.research.google.com/drive/19QlC6oNDTK5VMxvufI1uETQ7AXohlLDt?usp=sharing#scrollTo=6nZDXq48ZzTa">Data Visualization for Time Series</a></p><p><a href="https://medium.com/geekculture/greykite-the-new-forecasting-library-from-linkedin-case-bitcoin-price-prediction-981d21e79b31">Reference for GreyKite</a></p><h3>MIT License</h3><p>All code in this notebook is available as open source through the MIT license.</p><p>All text and images are free to use under the Creative Commons Attribution 3.0 license. <a href="https://creativecommons.org/licenses/by/3.0/us/">https://creativecommons.org/licenses/by/3.0/us/</a></p><p>These licenses let people distribute, remix, tweak, and build upon the work, even commercially, as long as they give credit for the original creation.</p><p>Copyright 2023 AI Skunks <a href="https://github.com/aiskunks">https://github.com/aiskunks</a></p><p>Permission is hereby granted, free of charge, to any person obtaining a copy of this software and associated documentation files (the “Software”), to deal in the Software without restriction, including without limitation the rights to use, copy, modify, merge, publish, distribute, sublicense, and/or sell copies of the Software, and to permit persons to whom the Software is furnished to do so, subject to the following conditions:</p><p>The above copyright notice and this permission notice shall be included in all copies or substantial portions of the Software.</p><p>THE SOFTWARE IS PROVIDED “AS IS”, WITHOUT WARRANTY OF ANY KIND, EXPRESS OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL THE AUTHORS OR COPYRIGHT HOLDERS BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE</p><img src="https://medium.com/_/stat?event=post.clientViewed&referrerSource=full_rss&postId=596eb7840cf9" width="1" height="1" alt="">]]></content:encoded>
        </item>
        <item>
            <title><![CDATA[Quiz on Statistical Inferences]]></title>
            <link>https://medium.com/@shreeshalini2912/quiz-on-statistical-inferences-9d42bcc0164a?source=rss-239affa7bb58------2</link>
            <guid isPermaLink="false">https://medium.com/p/9d42bcc0164a</guid>
            <category><![CDATA[statistical-inference]]></category>
            <category><![CDATA[quiz]]></category>
            <dc:creator><![CDATA[Shalini Shree]]></dc:creator>
            <pubDate>Tue, 14 Mar 2023 02:17:02 GMT</pubDate>
            <atom:updated>2023-03-14T02:17:02.749Z</atom:updated>
            <content:encoded><![CDATA[<ol><li>What is statistical inference?</li><li>What is a confidence interval?</li><li>What is the null hypothesis?</li><li>What is a p-value?</li><li>What does z-score signify?</li><li>What is the alternative hypothesis?</li><li>What is the mathematical formula for the chi-squared test?</li><li>What is a Type II error?</li><li>What is the significance level?</li><li>What is the central limit theorem?</li></ol><img src="https://medium.com/_/stat?event=post.clientViewed&referrerSource=full_rss&postId=9d42bcc0164a" width="1" height="1" alt="">]]></content:encoded>
        </item>
        <item>
            <title><![CDATA[Statistical Inference — Worked Example]]></title>
            <link>https://medium.com/@shreeshalini2912/statistical-inference-worked-example-33bc29c70032?source=rss-239affa7bb58------2</link>
            <guid isPermaLink="false">https://medium.com/p/33bc29c70032</guid>
            <category><![CDATA[statistical-inference]]></category>
            <category><![CDATA[data-science]]></category>
            <category><![CDATA[statistical-learning]]></category>
            <category><![CDATA[chi-square-test]]></category>
            <category><![CDATA[data-analysis]]></category>
            <dc:creator><![CDATA[Shalini Shree]]></dc:creator>
            <pubDate>Tue, 14 Mar 2023 02:08:26 GMT</pubDate>
            <atom:updated>2023-03-14T02:08:26.476Z</atom:updated>
            <content:encoded><![CDATA[<h3>Statistical Inference — Worked Example</h3><p><strong>The Chi-Squared Goodness of Fit Test</strong> is a statistical test used to determine whether a sample of categorical data comes from a population with a specific distribution. In other words, it tests whether the observed frequencies of the categories in the sample match the expected frequencies based on a specified distribution.</p><p>The test involves comparing the observed frequencies of each category in the sample to the expected frequencies based on the distribution being tested. The expected frequencies are calculated based on the distribution’s probabilities and the sample size. The differences between the observed and expected frequencies for each category are then squared, divided by the expected frequency, and added up to obtain a test statistic, which follows a chi-squared distribution.</p><p>If the test statistic is larger than the critical value from the chi-squared distribution for a given level of significance and degrees of freedom, we reject the null hypothesis that the sample comes from the specified distribution. In other words, we conclude that there is evidence that the sample does not follow the specified distribution.</p><p>The Chi-Squared Goodness of Fit Test is commonly used in fields such as biology, social sciences, and marketing research, where categorical data is frequently collected. It can be used to test hypotheses about the distribution of a population, such as whether it follows a normal distribution or a Poisson distribution.</p><h4>Create some demographic data for India and Nepal and run the chi-square goodness of fit test to see whether they differ or not:</h4><p>This dataset is a hypothetical representation of the religious composition of India and Nepal. It is represented as two separate pandas DataFrames, with each row representing an individual and their religion. The code uses the pandas “crosstab” function to create frequency tables for each country, which count the number of individuals in each religious group.</p><p>Specifically, the India DataFrame has 100,000 individuals who identify as Hindu, 60,000 individuals who identify as Muslim, 50,000 individuals who identify as Christian, 15,000 individuals who identify as Sikh, and 35,000 individuals who identify as &quot;other.&quot; The Nepal DataFrame has 60,000 individuals who identify as Hindu, 3,000 individuals who identify as Muslim, 250 individuals who identify as Christian, 75 individuals who identify as Sikh, and 150 individuals who identify as &quot;other.&quot;</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*2ryz8eBGqHR6WJmPelqruQ.png" /></figure><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*HIgUwFJcZepvc6uB9GHOUQ.png" /></figure><p>The chi-squared statistic value 84608.334271 represents how much the observed frequencies in Nepal deviate from the expected frequencies based on the population ratios in India. A larger value indicates a larger difference between the observed and expected frequencies, which suggests that the variable being studied i.e., Nepal may be associated with some other factor, such as geographic location or cultural differences, that is not accounted for in the population ratios.</p><img src="https://medium.com/_/stat?event=post.clientViewed&referrerSource=full_rss&postId=33bc29c70032" width="1" height="1" alt="">]]></content:encoded>
        </item>
        <item>
            <title><![CDATA[Statistical Inferences]]></title>
            <link>https://medium.com/aiskunks/statistical-inferences-db2bed5b8f86?source=rss-239affa7bb58------2</link>
            <guid isPermaLink="false">https://medium.com/p/db2bed5b8f86</guid>
            <category><![CDATA[statistics]]></category>
            <category><![CDATA[inferential-statistics]]></category>
            <category><![CDATA[data-science]]></category>
            <dc:creator><![CDATA[Shalini Shree]]></dc:creator>
            <pubDate>Mon, 13 Mar 2023 01:50:18 GMT</pubDate>
            <atom:updated>2023-03-27T16:26:52.639Z</atom:updated>
            <content:encoded><![CDATA[<blockquote>Shalini Shree, Nik Bear Brown</blockquote><figure><img alt="" src="https://cdn-images-1.medium.com/max/320/1*155_0fhtYUTOqmJB3jQqxA.jpeg" /></figure><h3>INTRODUCTION</h3><blockquote>LET’S BEGIN WITH AN EXAMPLE!</blockquote><p><em>Do you like Pizza?</em></p><p><em>If yes, ever wondered which kind of pizza people prefer?</em></p><p><em>NO, then let&#39;s try to figure it out.</em></p><p>Imagine you’re a pizza chef who wants to determine what toppings are most popular among your customers. You have a menu with 20 different toppings, and you want to know which ones are the most popular so that you can optimize your pizza-making process and inventory management.</p><p>However, surveying every single customer who orders pizza from your restaurant would be time-consuming and impractical. Instead, you decide to use sampling to estimate the popularity of each topping.</p><p>To do this, you randomly select a sample of 100 pizza orders from the past month. You record the toppings that were requested on each of these pizzas and count how many times each topping was ordered. This gives you a sample distribution of the toppings.</p><p>Based on this sample, you find that the most popular toppings are pepperoni, mushrooms, and sausage. You can use this information to make informed decisions about your inventory management, such as stocking up on more of these popular toppings and potentially reducing the inventory of less popular toppings.</p><p>However, it’s important to keep in mind that the sample of 100 pizza orders may not be an exact representation of all the pizza orders your restaurant receives. The sample may be biased in some way, such as only including orders from certain times of the day or certain days of the week. To account for this uncertainty, you can calculate a margin of error and a confidence interval to estimate the range of values in which the true popularity of each topping is likely to fall.</p><p>Sampling can be a fun and useful tool for making decisions in everyday life, such as optimizing your pizza-making process!</p><p><strong><em>Trying to find out the nature of a problem before making a decision using statistical tools on a sample of the entire population is referred to as Statistical Inference.</em></strong></p><p>Inferential statistics also allow for the quantification of uncertainty and the assessment of hypotheses about the population based on the sample data.</p><p><strong>Advantages of statistical inference</strong></p><ul><li>Enables making generalizations about a larger population based on a smaller sample of the population.</li><li>Allows for the quantification of uncertainty, so that the level of confidence in the results can be determined.</li><li>Helps to identify patterns and relationships in the data, and make predictions about future events.</li><li>Facilitates the testing of hypotheses and the comparison of multiple groups or variables.</li></ul><p><strong>Disadvantages of statistical inference</strong></p><ul><li>The conclusions drawn from statistical inference are only as reliable as the sample data. If the sample is not representative of the population, the results may be biased.</li><li>Inferential statistics can be affected by outliers and other anomalies in the data.</li><li>The choice of statistical methods and models used for inference can impact the results, so it is important to choose appropriate methods for the data and research questions.</li><li>The results of statistical inference are only an estimate and may not perfectly reflect the true population values.</li></ul><h3>SIZE OF THE SAMPLE?</h3><p>The appropriate sample size depends on several factors such as the variability of the population, the desired level of precision, the confidence level, and the research question being addressed. In general, a larger sample size is preferred as it provides a more accurate estimate of the population parameter with higher precision and reduces the sampling error.</p><p>However, it’s important to note that the sample size should not be too large, as it can lead to a waste of resources and time. Additionally, the sample should be representative of the population to ensure the generalisability of the results.</p><h3>TYPES OF SAMPLING TECHNIQUES</h3><h4><strong>Probability sampling methods</strong></h4><ol><li><strong><em>Simple Random Sampling:</em></strong><em> </em>Simple random sampling is the most common sampling technique used in data analysis. In this technique, every individual in the population has an equal chance of being selected. For example, if we are finding the preference of toppings on a pizza, we can randomly select a set of orders from the order history to see the most popular choices of toppings in those orders.</li></ol><p><strong><em>Advantages</em></strong><em>:</em></p><ul><li>It is easy to implement.</li><li>It is unbiased, as every individual has an equal chance of being selected.</li></ul><p>It is unbiased, as every individual has an equal chance of being selected.</p><p><strong><em>Disadvantages</em></strong><em>:</em></p><ul><li>It may not be representative of the population if the sample size is small.</li><li>It can be expensive and time-consuming if the population is large.</li></ul><p><strong><em>2. Systematic sampling:</em></strong> Systematic sampling is similar to simple random sampling, but it is usually slightly easier to conduct. Every member of the population is listed with a number, but instead of randomly generating numbers, individuals are chosen at regular intervals.</p><p><strong><em>3. Stratified Sampling:</em></strong> Stratified sampling is a sampling technique used when the population can be divided into distinct subgroups or strata. In this technique, individuals are randomly selected from each stratum in proportion to their size. This ensures that the sample is representative of the population.</p><p><strong><em>Advantages</em></strong><em>:</em></p><ul><li>It ensures that the sample is representative of the population.</li><li>It reduces the sampling error, as individuals are selected from each stratum.</li></ul><p><strong><em>Disadvantages</em></strong><em>:</em></p><ul><li>It can be complex to implement if the population has many strata.</li><li>It may not be effective if the population is not well-defined into strata.</li></ul><p><strong><em>4. Cluster Sampling:</em></strong> Cluster sampling is a technique used when it is difficult or expensive to obtain a sample from the population. In this technique, the population is divided into clusters, and a random sample of clusters is selected. Then, all individuals within the selected clusters are included in the sample.</p><p><strong><em>Advantages</em></strong><em>:</em></p><ul><li>It is cost-effective, as it reduces the time and cost of sampling.</li><li>It can be effective if the population is geographically dispersed.</li></ul><p><strong><em>Disadvantages</em></strong><em>:</em></p><ul><li>It may not be representative of the population if the clusters are not well-defined.</li><li>It may introduce bias if the clusters are not homogenous.</li></ul><h4><strong>NON PROBABILITY SAMPLING METHOD</strong></h4><ol><li><strong><em>Convenience Sampling:</em></strong> Convenience sampling is a type of non-probability sampling technique in which the sample is chosen based on the ease of access to the individuals in the population. Convenience sampling is easy to use and can be less time-consuming than other sampling techniques. However, convenience sampling is not representative of the population, as individuals who are more accessible may not be representative of the entire population.</li><li><strong><em>Quota Sampling:</em></strong> Quota sampling is a type of non-probability sampling technique in which the sample is chosen based on pre-defined characteristics such as age, gender, or income. The population is divided into subgroups based on these characteristics, and a quota is set for each subgroup. The sample is then chosen from each subgroup until the quota is met. Quota sampling is relatively easy to use and can be less expensive than other sampling techniques. However, quota sampling may not be representative of the entire population, as individuals who are not part of the pre-defined characteristics may be excluded from the sample.</li><li><strong><em>Snowball Sampling:</em></strong> Snowball sampling is a type of non-probability sampling technique in which the initial participants in the sample are chosen based on their knowledge of other individuals in the population who have similar characteristics. The initial participants are asked to identify other individuals in the population who meet the criteria for the study, and those individuals are then asked to participate. Snowball sampling is useful when the population is difficult to access or when it is difficult to identify the entire population. However, snowball sampling may introduce bias into the sample, as individuals who are more connected or more willing to participate may be overrepresented in the sample.</li><li><strong><em>Purposive Sampling:</em></strong> Purposive sampling is a type of non-probability sampling technique in which the sample is chosen based on the researcher’s judgment or purpose of the study. The researcher selects individuals who are likely to provide the necessary information or represent the characteristics of interest in the study. Purposive sampling can be useful when the population is difficult to access or when the research question requires a specific type of participant. However, purposive sampling may introduce bias into the sample, as the researcher’s judgment may be subjective and influenced by personal biases.</li></ol><h3>WHAT IS BIAS?</h3><p>Bias refers to any systematic error or deviation from the true value that affects the accuracy and validity of the results. Bias can arise from various sources, such as the sampling method, the measurement instrument, the data collection process, or the data analysis technique used.</p><p>Some common types of bias include:</p><p><strong>Sampling Bias:</strong> Sampling bias occurs when the sample selected for the study is not representative of the entire population or when certain groups are over- or underrepresented in the sample. Sampling bias can lead to inaccurate or misleading results that do not generalize to the population.</p><p><strong>Measurement Bias:</strong> Measurement bias occurs when the measurement instrument used to collect data systematically under- or overestimates the true value of the variable of interest. Measurement bias can lead to inaccurate or unreliable data that do not reflect the true state of the phenomenon.</p><p><strong>Response Bias:</strong> Response bias occurs when the participants in the study do not provide accurate or truthful responses to the survey questions or interview prompts. Response bias can arise from social desirability bias, where participants give answers that they believe are socially acceptable or desirable, or from recall bias, where participants have difficulty recalling past events or experiences accurately.</p><p><strong>Confounding Bias:</strong> Confounding bias occurs when the relationship between two variables is affected by a third variable that is not accounted for in the analysis. Confounding bias can lead to spurious or misleading results that do not reflect the true association between the variables of interest.</p><p><strong>Selection Bias:</strong> Selection bias occurs when individuals or groups are selected for inclusion in the study based on non-random criteria, such as availability, convenience, or personal preference. Selection bias can lead to a biased sample that does not represent the population of interest.</p><h3>Types of statistical inferences:</h3><p><strong>Deductive inference:</strong> Also known as analytic or theoretical inference, this involves deducing conclusions based on prior knowledge or assumptions about a population, without the need for collecting sample data.</p><p><strong>Inductive inference:</strong> Also known as empirical or observational inference, this involves drawing conclusions about a population based on a sample of data collected from that population.</p><h4>Deductive Inference</h4><p>A sort of statistical inference known as deductive inference involves drawing inferences about a population based on knowledge or presumptions about it without the requirement for sample data. Instead of using sample data to draw conclusions about the population, this sort of inference is founded on logical reasoning and prior knowledge.</p><p>There are several techniques used in deductive inference, including:</p><ul><li><em>Theoretical modelling:</em> This entails utilizing equations and mathematical models to portray the connections and interactions among members of a population. Based on the underlying assumptions and inputs, the models may be used to generate inferences and predictions about the population.</li><li><em>Formal logic:</em> This implies drawing conclusions based on presumptions or past information by applying formal logic and reasoning. Formal logic may be applied to draw conclusions regarding the connections between variables and to generate predictions based on those connections.</li><li><em>Bayesian inference:</em> This includes combining sample data and previous information to draw conclusions about a population using the Bayesian probability theory. With the use of fresh evidence, Bayesian inference may be used to revise prior assumptions about a population and produce population-specific predictions.</li></ul><h4>Inductive inference</h4><p>Inductive inference involves making inferences about a population based on a sample of data. This kind of inference is predicated on the notion that conclusions about a wider population may be drawn from the characteristics of a sample.</p><ul><li><em>Point estimation:</em> This involves using sample data to estimate the value of a population parameter, such as the mean or variance. Point estimation techniques include methods such as the sample mean and maximum likelihood estimation.</li><li><em>Confidence intervals:</em> This involves using sample data to construct an interval that is believed to contain the true value of a population parameter with a certain level of confidence. Confidence intervals can be used to estimate the range of values that a population parameter is likely to take.</li><li><em>Hypothesis testing:</em> This involves using sample data to test a hypothesis about a population parameter. Hypothesis testing involves making a null hypothesis (e.g., the population’s mean is equal to a certain value) and an alternative hypothesis and then using sample data to determine which hypothesis is more likely to be true.</li><li><em>Regression analysis:</em> This involves using sample data to fit a regression model that describes the relationship between one or more independent variables and a dependent variable. Regression analysis can be used to make predictions about the dependent variable based on the values of the independent variables.</li><li><em>Bayesian inference:</em> This involves using Bayesian probability theory to make inferences about a population based on prior knowledge and sample data. Bayesian inference can be used to update prior beliefs about a population based on new data, and to make predictions about the population based on that updated information.</li></ul><h3>Basic Probability</h3><p>Probability can roughly be described as the chance of an event or sequence of events occurring.</p><p>Experiment — are uncertain situations, which could have multiple outcomes. A coin toss is an experiment.</p><p>Outcome is the result of a single trial. So, if head lands, the outcome of or coin toss experiment is “Heads”</p><p>Event is one or more outcomes from an experiment. “Tails” is one of the possible events for this experiment.</p><ul><li>The Probability that an event occurs with certainty is 1</li><li>The Probability that an event will not occur surely is 0</li><li>The Probability of the complement of an event is 1 minus the probability of that event.</li><li>The probability of at least 1 of 2 (or more) things that can not simultaneously occur (mutually exclusive) is the sum of their respective probabilities</li></ul><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*eXwQ1dZshwZvG6KptD1ZDQ.png" /></figure><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*HYtWhE0dBt6deQ5zcd5RYw.png" /></figure><h3>Conditional Probability</h3><p>conditional probability is the probability of event A happening, given that we have some information about the occurrence of event B.</p><p>The conditional probability of A given B is denoted by P(A|B) and is calculated as follows:</p><p>P(A|B) = P(A and B) / P(B)</p><p>where P(A and B) is the probability of both A and B occurring, and P(B) is the probability of B occurring.</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*gm48uq-t31wJ0wqgcLP9Ag.png" /></figure><h3>Probability Distribution</h3><p>A probability distribution is a statistical function that describes the likelihood of obtaining all possible values that a random variable can take.</p><p>If our random variable has discrete values, the probability distribution is the <strong>probability mass function (PMF)</strong> for that variable; similarly, if our random variable has continuous values, the distribution is termed a <strong>probability density function (PDF)</strong>.</p><p>One of the most common Probability Distribution Functions is the Normal Distribution.</p><h4>Normal Distribution</h4><p>The normal distribution, also known as the Gaussian distribution, is a symmetric probability distribution about the mean, indicating that data near the mean occur more frequently than data distant from the mean. The normal distribution will show as a bell curve on a graph.</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*o5ZQ8FMBJ5hpEF-hEtu9vg.png" /><figcaption>The graph shows that the distribution of target value is relatively flat or uniform over that range of values. Also, indicates that the PDF is approximately constant over that range.</figcaption></figure><h3>Z-score</h3><p>A z-score (also called a standard score) is a statistical measure that indicates how many standard deviations a data point is away from the mean of the distribution. It is calculated by subtracting the mean from the data point and then dividing it by the standard deviation.</p><p>z-score = (x — μ) / σ</p><p>E.g.: Now suppose we want to find the z-score for a particular pizza topping, pepperoni. We can do this by first calculating the mean and standard deviation of the number of pepperoni slices consumed by each person. Let’s assume that the data shows that, on average, each person consumes 2 slices of pepperoni pizza, with a standard deviation of 0.5 slices.</p><p>To find the z-score for a person who ate 3 slices of pepperoni pizza. We can plug the values into the formula as follows:</p><p>z-score = (3–2) / 0.5 = 2</p><p>This means that the person who ate 3 slices of pepperoni pizza is 2 standard deviations above the mean number of pepperoni slices consumed by a person.</p><h3>Inference from Stats</h3><h4>Sample Mean and Population Mean</h4><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*13CfVaWno1yPAN2skPoP0Q.png" /></figure><h3>Central Limit Theorem</h3><p>The central limit theorem (CLT) is a statistical theory that claims that given a big enough sample size from a population with a limited degree of variance, the mean of all samples from the same population will be nearly equal to the overall mean. Furthermore, all of the samples will follow an almost normal distribution pattern, with their variances roughly equal to the variance of the population divided by the size of each sample.</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*0Izl5HJ7zfzIUUoYnng8NQ.png" /></figure><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*m5JTO011s7M2WsXeHq2MoA.png" /></figure><figure><img alt="" src="https://cdn-images-1.medium.com/max/888/1*tQG4-se-Fka8vvKflnyZ0Q.png" /></figure><p>In the graphs above, the yellow curve is the predicted Gaussian distribution from the Central Limit Thereom. Notice that the rate of convergence of the sample mean to the Gaussian depends on the original parent distribution. Also, the mean of the Gaussian distribution is the same as the original parent distribution, the width of the Gaussian distribution varies with sample size as 1/√n.</p><h3>Confidence Interval</h3><p>Confidence Interval (CI) is a form of estimate derived from observable data statistics. This suggests a range of possible values for an unknown parameter (for example, the mean). The interval is accompanied by a level of confidence that the real parameter is within the recommended range.</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*YZekdFOWqQSp7iqPaGBFag.png" /></figure><p>z-critical value: 1.6448536269514722 — This is the z-critical value for a 95% confidence interval. It is obtained from the standard normal distribution using the ppf (percent point function) method.</p><p>A confidence interval of 95% would mean that if we take many samples and create confidence intervals for each of them, 95% of our samples’ confidence intervals will contain the true population mean.</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*s0e_60eieiM89Njl2PO5Zw.png" /></figure><figure><img alt="" src="https://cdn-images-1.medium.com/max/849/1*oRCbLQXapZ82I4Nn2dPICw.png" /></figure><p>It is easily visible that 95% of the time the blue lines(the sample mean) overlap with the red line(the true mean), also 5% of the time it is expected to not overlap with the red line(the true mean).</p><p>The plot shows that the sample means for each trial tend to cluster around the true population mean, and the confidence intervals become narrower as the number of trials increases. This is consistent with the Central Limit Theorem, which states that as the sample size increases, the distribution of sample means approaches a normal distribution with a mean equal to the true population mean and a standard deviation equal to the standard deviation of the population divided by the square root of the sample size.</p><h3>Hypothesis Testing</h3><p>Hypothesis testing, also known as confirmatory data analysis, is a test that may be performed by observing a process that is described by a set of random variables. A statistical hypothesis test is a statistical inference procedure.</p><h4>NULL Hypothesis</h4><p>The Null Hypothesis (𝐻0) is a generic statement or default stance in inferential statistics that there is no link between two measurable phenomena or no correlation among groups.</p><p>Statistical hypothesis tests are based on a statement known as the null hypothesis, which asserts that nothing meaningful is happening between the variables being tested.</p><p>Therefore, in our case, the Null Hypothesis would be: <strong>The frequency of Pepperoni topping is not different from the frequency of other toppings.</strong></p><h4>Alternate Hypothesis</h4><p>The alternate hypothesis (𝐻1) is just an alternative to the null hypothesis.</p><p><strong>The Null Hypothesis is considered to be true, and the statistical proof is required to reject it if in favor of an Alternative Hypothesis.</strong></p><h4>Type 1 and Type 2 Error</h4><p>In statistical hypothesis testing, a type I error is the rejection of a true null hypothesis, while a type II error is the non-rejection of a false null hypothesis</p><h4>P-Value</h4><p>The p-value, also known as the probability value, is the likelihood of receiving test findings that are at least as severe as the outcomes actually observed during the test, provided that the null hypothesis is valid.</p><p>Assume we’ve set a significance level of () = 0.05.</p><p>This indicates that if the p-value is less than 0.05, we reject our Null hypothesis and accept the Alternative as true.</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*ksvn0jNyzo8bxV-XMljtMQ.png" /></figure><p>Here, the z-statistic is 0.7767, which indicates that the sample mean of the Pepperoni pizza orders is 0.7767 standard deviations above the population mean. The p-value is 0.43731, which is greater than the typical significance level of 0.05. This means that we fail to reject the null hypothesis and conclude that there is insufficient evidence to suggest that the mean order frequency of Pepperoni pizza is significantly different from the overall mean order frequency of all pizza toppings.</p><h4>Gosset’s / Student’s t-test</h4><p>The T-test is a statistical test used to detect if a numerical data sample differs significantly from the population or whether two samples differ from one another.</p><p>A t-test addresses this difficulty and allows us to conduct a hypothesis test on a lower sample size.</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*4WzjCA1D7fFZPD7RqD7G2g.png" /><figcaption>Since the p-value is greater than the common significance level of 0.05, we fail to reject the null hypothesis</figcaption></figure><h3>Chi-Square Test</h3><p>The Chi-Square test is a statistical hypothesis test that is used to determine whether there is a significant association between two categorical variables. It is used to test the independence of two variables.</p><p>The Chi-Square test requires the following assumptions:</p><ul><li>The data are categorical.</li><li>The observations are independent.</li><li>The expected frequency for each category is at least 5.</li></ul><p><strong>The chi-Squared Goodness of fit Test</strong> is a statistical test used to determine whether a sample of categorical data follows a specified distribution. In other words, it helps us to determine whether the observed frequencies of a categorical variable are significantly different from the expected frequencies.</p><p>The formula for the Chi-Squared test:</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/384/1*pMrY-N_MlscgEajfBUW9QQ.png" /></figure><p>where,</p><p>O = each observed value,</p><p>E = each expected value.</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*KWDiD8hZ53jZ1FMoHzBp_A.png" /></figure><p>In this case, the chi-squared test statistic is 180.65, and the p-value is 0.472, which is greater than the significance level of 0.05. Therefore, we fail to reject the null hypothesis and conclude that there is no significant association between the price of a pizza and its frequency of purchase.</p><h3>REFERENCES</h3><ul><li><a href="https://www.w3schools.com/statistics/statistics_statistical_inference.php">https://www.w3schools.com/statistics/statistics_statistical_inference.pp</a></li><li>Introduction to statistical Inferences — O’Reilly</li><li><a href="https://soc.utah.edu/sociology3112/chi-square.php">https://soc.utah.edu/sociology3112/chi-square.php</a></li></ul><h3>MIT License</h3><p>All code in this notebook is available as open source through the MIT license.</p><p>All text and images are free to use under the Creative Commons Attribution 3.0 license. <a href="https://creativecommons.org/licenses/by/3.0/us/">https://creativecommons.org/licenses/by/3.0/us/</a></p><p>These licenses let people distribute, remix, tweak, and build upon the work, even commercially, as long as they give credit for the original creation.</p><p>Copyright 2023 AI Skunks <a href="https://github.com/aiskunks">https://github.com/aiskunks</a></p><p>Permission is hereby granted, free of charge, to any person obtaining a copy of this software and associated documentation files (the “Software”), to deal in the Software without restriction, including without limitation the rights to use, copy, modify, merge, publish, distribute, sublicense, and/or sell copies of the Software, and to permit persons to whom the Software is furnished to do so, subject to the following conditions:</p><p>The above copyright notice and this permission notice shall be included in all copies or substantial portions of the Software.</p><p>THE SOFTWARE IS PROVIDED “AS IS”, WITHOUT WARRANTY OF ANY KIND, EXPRESS OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL THE AUTHORS OR COPYRIGHT HOLDERS BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE</p><img src="https://medium.com/_/stat?event=post.clientViewed&referrerSource=full_rss&postId=db2bed5b8f86" width="1" height="1" alt=""><hr><p><a href="https://medium.com/aiskunks/statistical-inferences-db2bed5b8f86">Statistical Inferences</a> was originally published in <a href="https://medium.com/aiskunks">AI Skunks</a> on Medium, where people are continuing the conversation by highlighting and responding to this story.</p>]]></content:encoded>
        </item>
    </channel>
</rss>