<?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 Aliakbar Rzayev on Medium]]></title>
        <description><![CDATA[Stories by Aliakbar Rzayev on Medium]]></description>
        <link>https://medium.com/@raliakbar27?source=rss-d84985d74b80------2</link>
        <image>
            <url>https://cdn-images-1.medium.com/fit/c/150/150/1*WSORE27toYFNM9RwYXL5dw.jpeg</url>
            <title>Stories by Aliakbar Rzayev on Medium</title>
            <link>https://medium.com/@raliakbar27?source=rss-d84985d74b80------2</link>
        </image>
        <generator>Medium</generator>
        <lastBuildDate>Sat, 21 Nov 2026 09:31:18 GMT</lastBuildDate>
        <atom:link href="https://medium.com/@raliakbar27/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[Understanding and Applying Hypothesis Tests]]></title>
            <link>https://raliakbar27.medium.com/understanding-and-applying-hypothesis-tests-c8bac0a09fc5?source=rss-d84985d74b80------2</link>
            <guid isPermaLink="false">https://medium.com/p/c8bac0a09fc5</guid>
            <category><![CDATA[sampling]]></category>
            <category><![CDATA[hypothesis-testing]]></category>
            <category><![CDATA[a-b-testing]]></category>
            <category><![CDATA[correlation]]></category>
            <category><![CDATA[descriptive-statistics]]></category>
            <dc:creator><![CDATA[Aliakbar Rzayev]]></dc:creator>
            <pubDate>Tue, 18 Jul 2023 09:32:47 GMT</pubDate>
            <atom:updated>2023-07-26T06:58:03.148Z</atom:updated>
            <content:encoded><![CDATA[<figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/0*kjEO-JI4sdZw8xex" /></figure><p>Hypothesis testing is a statistical method used to determine if there is enough evidence to support a claim about a population based on sample data. The goal is to draw conclusions and make decisions based on statistical evidence and determine whether observed differences are significant or simply due to <strong>random chance</strong>. In this article, we will explore some key statistical concepts and learn how to apply them effectively.</p><h3>Sampling:</h3><figure><img alt="" src="https://cdn-images-1.medium.com/max/810/1*UhYnLF3CCLytITJtMHUcWw.png" /><figcaption><a href="https://www.investopedia.com/terms/s/sampling.asp">https://www.investopedia.com/terms/s/sampling.asp</a></figcaption></figure><p>Sampling is the process of selecting a subset from a larger population that is assumed to represent the characteristics of the entire population. When working with samples, we aim to make generalizations about the entire population using less data.</p><p>For instance, in the future of AI, we might rely on smaller but well-representative samples to make predictions and inferences, instead of gathering massive amounts of data.</p><h3>Descriptive Statistics:</h3><figure><img alt="" src="https://cdn-images-1.medium.com/max/802/1*ULIBS_ecqVoXyOwZu-pp0A.png" /><figcaption><a href="https://www.investopedia.com/terms/d/descriptive_statistics.asp">https://www.investopedia.com/terms/d/descriptive_statistics.asp</a></figcaption></figure><p>Descriptive statistics involves summarizing and describing the characteristics of a dataset. When dealing with skewed data or outliers, using the median instead of the mean can provide a more accurate representation of the data.</p><h3>Confidence Intervals:</h3><figure><img alt="" src="https://cdn-images-1.medium.com/max/789/1*2j3ory870To-9zh92zdXtg.png" /><figcaption><a href="https://www.investopedia.com/terms/c/confidenceinterval.asp">https://www.investopedia.com/terms/c/confidenceinterval.asp</a></figcaption></figure><p>Confidence intervals are ranges that are likely to contain the true value of a population parameter. Instead of relying on single point estimates, confidence intervals offer richer information about the uncertainty of our estimates.</p><h3>Correlation:</h3><figure><img alt="" src="https://cdn-images-1.medium.com/max/808/1*qzl0bTuqDYk0KDeHqinBBg.png" /><figcaption><a href="https://www.investopedia.com/terms/c/correlation.asp">https://www.investopedia.com/terms/c/correlation.asp</a></figcaption></figure><p>Correlation measures the relationship, direction, and strength between two variables. It provides valuable insights into how changes in one variable affect another.</p><h3>A/B Testing:</h3><figure><img alt="" src="https://cdn-images-1.medium.com/max/742/1*O7I4iVQdIaAxl5u0VB32FA.png" /><figcaption><a href="https://www.investopedia.com/terms/a/a-b-split.asp">https://www.investopedia.com/terms/a/a-b-split.asp</a></figcaption></figure><p>And we will talk about the specific type of Hypothesis testing: A/B testing, it helps determine if there is a significant difference between two groups.</p><p><strong>To conduct hypothesis testing, we typically follow these steps:</strong></p><ol><li>Formulate the null and alternative hypotheses.</li><li>Check the assumptions of the selected tests (Normality ssumption, Variance Homogeneity)</li><li>Apply the appropriate statistical test based on the assumptions and nature of the data.</li><li>Conclusion and comment according to the p-value.</li></ol><p>We commonly use a significance level (alpha) of 0.05, which means we aim to achieve a p-value less than 0.05 to reject the null hypothesis(H0). On the other hand, if the p-value is greater than 0.05 we will fail to reject the null hypothesis. <strong>There is no such thing in theory;</strong> <em>accepting the alternative hypothesis(H1).</em></p><h3>Two Sample T-test</h3><p>After formulating the Hypotheses, We should check the assumptions:</p><blockquote>Normality Assumption:</blockquote><ul><li>Normality assumption will be checked with the<strong> Shapiro Test</strong>. If the value of p is less than 0.05, we will reject the null hypothesis and it means that Normality is not provided.</li></ul><blockquote>Variance Homogeneity:</blockquote><ul><li><strong>Levence Test </strong>will be used to examine this homogeneity. Again checking the p-value to see if we will reject or fail to reject the null hypothesis, in order to make a decision whether the Variance Homogeneity provided or not.</li></ul><p>If <strong>assumptions are met</strong>, we will use the <em>Independent Two-Sample T-test</em> (parametric test). If the <strong>assumptions are not met</strong>, we will go for non-parametric test the <em>Mann-Whitney U test</em>.</p><p><strong>Note</strong>: If the Normality assumption is not met, we can directly use <em>Mann-Whitney U test </em>without checking the Variance homogeneity. On the other hand, if Normality provided but Homogeneity is not met, we can use <em>Independent Two-Sample T-test </em>and<em> </em>giving an<em> </em>argument that is entered as False.</p><h3>Two Sample Ratio-test</h3><p>When dealing with proportions or percentages, such as success rates, we use the two-sample proportion test to compare the proportions of two groups. It is commonly employed when dealing with categorical data or success/failure outcomes.</p><p>In this test we will directly use <em>Proportion Z-test</em> to get the p-value in order to decide about the Hypotheses. Unlike the T-test we will use number of successes and observations for the input of test function.</p><h3>ANOVA — Analysis of Variance</h3><p>If we want to compare the means of more than two groups simultaneously, we use ANOVA. It is particularly useful when we want to investigate if there are significant differences between the groups based on a single continuous dependent variable.</p><p>Like in A/B testing, we need to check assumptions such as normality and variance homogeneity before applying ANOVA.</p><p>If assumptions are met, we use <em>one-way ANOVA</em>. If not, we use the <em>Kruskal-Wallis test</em>, a non-parametric alternative.</p><h3>Conclusion</h3><p>In conclusion, A/B testing and the statistical concepts associated with it play a crucial role in data science. By understanding these concepts and applying the appropriate tests, we can draw meaningful conclusions from our data and make informed decisions. It is essential to check assumptions and choose the right test for the specific context to ensure the validity and reliability of our results.</p><p><strong>References</strong></p><ul><li>Miuul</li><li>Investopedia</li></ul><img src="https://medium.com/_/stat?event=post.clientViewed&referrerSource=full_rss&postId=c8bac0a09fc5" width="1" height="1" alt="">]]></content:encoded>
        </item>
    </channel>
</rss>