365 sql just exploring

Author: ptyadana

SQL for Data Science Interviews - 365 DS/01_mock_interview.sql

@@ -0,0 +1,106 @@
+/*
+Mock Interview 1)
+
+-- table name: post_events
+
+-- user_id int
+-- created_at datetime
+-- event_name varchar
+-- (event, post, cancel)
+
+
+Question: What information would you like to start off by pulling to get an overall understanding of 
+the post feature?
+
+
+Possible Answer:
+We might want to get an idea of OVERALL HEALTH. 
++ Total number of posts (number of enters)
++ Posts Made by Date
++ Success Rate
++ Cancel Rate
+
+*/
+
+
+/***** Success Rate *****/
+-- success rate by date
+-- date | success rate = number of posts / number of enters
+
+-- explaination in english
+-- group by date
+-- count of number of posts / count of number enters
+
+SELECT created_at, 
+	COUNT(CASE WHEN event_name = 'post' 1 ELSE null END) * 1.00 / 
+	COUNT (CASE WHEN event_name = 'enter' 1 ELSE null END) * 100 AS percent_success
+FROM post_events
+GROUP BY created_at
+ORDER BY created_at;
+
+
+/*
+Question: When sucess rates are low, how can we diagonis the issue?
+
+Possible Answer: There can be several approach to this problem. But one possible way is we can take a look at the map
+out between created_at and success_rate. And see whether there is a pattern during a certain period of time.
+This can be one off dip, etc. Based on this information, we can take a further look at if there is any underlying
+issue in the application or not. Or if there is any potential group of users who are causing this kind of unsccessful posts.
+*/
+
+
+/*
+Questions: 
+
+What are the success rates by day?
+
+Which day of the week has the lowest success rate?
+*/
+
+-- group by dow of created_date
+-- average of perc_success
+-- order by per_success
+-- day | per_sucess 
+
+WITH created_events AS(
+	SELECT created_at, 
+		COUNT(CASE WHEN event_name = 'post' 1 ELSE null END) * 1.00 / 
+		COUNT (CASE WHEN event_name = 'enter' 1 ELSE null END) * 100 AS percent_success
+	FROM post_events
+	GROUP BY created_at
+	ORDER BY created_at)
+
+SELECT EXTRACT (dow FROM created_at) AS dow,
+AVG(percent_success)
+FROM created_events
+GROUP BY 1
+ORDER BY 2 ASC;
+
+
+/*
+Question: What could be a problem if we're aggregating on percent success?
+
+Possible Answer: this can lead to a problem that we're not taking into consideration of underlying distribution
+of percent success across the dates.
+*/
+
+
+SELECT EXTRACT (dow FROM created_at) AS dow,
+	COUNT(CASE WHEN event_name = 'post' 1 ELSE null END) * 1.00 / 
+	COUNT (CASE WHEN event_name = 'enter' 1 ELSE null END) * 100 AS percent_success
+FROM created_events
+GROUP BY 1
+ORDER BY 2 ASC;
+
+/*************************** Using actual table *******************************/
+
+SELECT *
+FROM interviews.post_events
+LIMIT 10;
+
+SELECT EXTRACT (dow FROM created_at) AS dow,
+	COUNT(CASE WHEN event_name = 'post' THEN 1 ELSE null END) * 1.00 / 
+	COUNT (CASE WHEN event_name = 'enter' THEN 1 ELSE null END) * 100 AS percent_success
+FROM interviews.post_events
+GROUP BY 1
+ORDER BY 2 ASC;

SQL for Data Science Interviews - 365 DS/02_mock_interview.sql

@@ -0,0 +1,50 @@
+/*
+Mock Interview 2)
+
+Question:
+
+-- 1) Find the date with the highest total energy consumption from our datacenters.
+-- 2) Output the date along with the total engery consumption across all datacenters.
+
+-- Table: eu_energy
+date			datetime
+consumption		int
+	
+-- Table: asia_energy
+date			datetime
+consumption		int
+
+-- Table: na_energy
+date			datetime
+consumption		int
+
+*/
+
+/******
+
+Always make sure you under the tables and columns correctly. Ask the interviewer if you need to make any assumptions
+on the columns data, etc.
+
+1) So there are 3 tables representing energy consumptions across different continents. 
+Can I assume that there is only one energy consumption for each particular date Or can there be a multiple consumptions
+for a specific date? Is there any possible missing values too?
+
+We can just sum it up across different dates across the different tables.
+*****/
+
+
+/*
+Question: What would you do if in the first table there are two of the same dates with
+different energy consumptions?
+
+Possibe Answer: 
+Then we can just group by using Date and sum the engery consumptions.
+*/
+
+
+/****
+
+Clarification Question back to Interviwer:
+Is there a situation that there are Multiple Dates with the same highest Total Energy Consumption?
+****/
+