Skip to main content

Central Limit Theorem and Hypothesis Testing

Another post starts with you beautiful people!
Today we will learn about an important topic related to statistics.

Statistical inference is the process of deducing properties of an underlying distribution by analysis of data. Inferential statistical analysis infers properties about a population: this includes testing hypotheses and deriving estimates.
Statistics are helpful in analyzing most collections of data. Hypothesis testing can justify conclusions even when no scientific theory exists.

You can find more about this here- tell me more about Statistical_hypothesis_testing

Here our case study will be Average Experience of Data Science Specialization(DSS) batch taught in a leading University with Statistical Inference.
We will aim to study how accurately can we characterize the actual average participant experience (population mean) from the samples of data (sample mean). We can quantify the certainty of outcome through the confidence intervals.


Let's plot the distribution of experience-

Result:-

Estimating DSS Experience from samples:-

We are now drawing samples for 1000 times (NUM_TRIALS) and compute the mean each time. The distribution is plotted to identify range of values it can take. The original data has experience raging between 0 years and 20 years and spread across it.
Result:-


The above plot is histogram of mean of samples for any given n. 
As you vary n from 1 to 128 you will see the following summary for the Sampling Distribution of Mean chart-
n = 1, Mean = 10.301, Std Dev = 5.689, 5% Pct = 1.000, 95% Pct = 19.000
n = 2, Mean = 10.431, Std Dev = 3.984, 5% Pct = 4.000, 95% Pct = 17.000
n = 4, Mean = 10.388, Std Dev = 2.878, 5% Pct = 5.500, 95% Pct = 15.000
n = 8, Mean = 10.407, Std Dev = 2.031, 5% Pct = 7.000, 95% Pct = 13.875
n = 10, Mean = 10.455, Std Dev = 1.794, 5% Pct = 7.600, 95% Pct = 13.405
n = 16, Mean = 10.467, Std Dev = 1.372, 5% Pct = 8.188, 95% Pct = 12.688
n = 32, Mean = 10.470, Std Dev = 0.984, 5% Pct = 8.844, 95% Pct = 12.125
n = 64, Mean = 10.446, Std Dev = 0.693, 5% Pct = 9.297, 95% Pct = 11.609

n = 128, Mean = 10.451, Std Dev = 0.502, 5% Pct = 9.625, 95% Pct = 11.258

One important point is to note the trend in values of Std Dev in the above table as n increases.

As n increases the Std Dev of Sampling distribution of Mean reduces from 5.689 (n = 1) to 0.502 (n = 128). 
As n increases the Sampling distribution of Mean looks more normal. 
It is observed that when n >= 30 irrespective of the original distribution the Sampling distribution of Mean becomes normal distribution.

Definition of Central Limit Theorem:-
In probability theory, the central limit theorem (CLT) establishes that, for the most commonly studied scenarios, when independent random variables are added, their sum tends toward a normal distribution (commonly known as a bell curve) even if the original variables themselves are not normally distributed [Ref: Wikipedia]


The key point to note is that irrespective of the original distribution then sum (and hence average as well) of large samples tend towards normal distribution. This gives power to estimate the confidence intervals.

Normal distribution has following characteristics:_


Function to check if the true mean lies within 90% Confidence Interval:-

Getting the average experience estimate from sample and Confidence Intervals:-
Now given the sample size n we can estimate the sample mean and confidence interval. The confidence interval is estimated assuming normal distribution which really holds good when n >= 30.

When n is increased the confidence interval becomes smaller which implies that results are obtained with higher certainty.

Execute the code below multiple times and check how often the population mean of 10.435 will lie within 90% confidence interval. It should be on average 9 out of 10 times i.e 90%-

Hypothesis Testing(HT):-
Let us define the Hypotheses as follows:

  • H0 : Average Experience of Current Batch & Previous batch are same
  • H1 : Average Experience of Current Batch & Previous batch are different
Repeat the same steps for previous batch-




Result:-


Below code describes to do hypothesis testing based on the intuition from Sampling distribution of mean. There will be differences with exact math based on various conditions and assumptions made.

Show the frequency distribution of experience:-

Code for dumping the experiment results to a csv file:-

Demonstrate how the standard deviation of sample mean reduce and the distribution goes closer to normal distribution:-


Stats functions for Hypothesis Testing taken from here- hypothesis-testing-comparing-two-groups


Comments


  1. After reading this blog i very strong in this topics and this blog really helpful to all.Data Science online Course Bangalore



    ReplyDelete

Post a Comment

Popular posts from this blog

How to install and compile YOLO v4 with GPU enable settings in Windows 10?

Another post starts with you beautiful people! Last year I had shared a post about  installing and compiling Darknet YOLOv3   in your Windows machine and also how to detect an object using  YOLOv3 with Keras . This year on April' 2020 the fourth generation of YOLO has arrived and since then I was curious to use this as soon as possible. Due to my project (built on YOLOv3 :)) work I could not find a chance to check this latest release. Today I got some relief and successfully able to install and compile YOLOv4 in my machine. In this post I am going to share a single shot way to do the same in your Windows 10 machine. If your machine does not have GPU then you can follow my  previous post  by just replacing YOLOv3 related files with YOLOv4 files. For GPU having Windows machine, follow my steps to avoid any issue while building the Darknet repository. My machine has following configurations: Windows 10 64 bit Intel Core i7 16 GB RAM NVIDIA GeForce G...

How can I make a simple ChatBot?

Another post starts with you beautiful people! It has been a long time of posting a new post. But my friends in this period I was not sitting  where I got a chance to work with chatbot and classification related machine learning problem. So in this post I am going to share all about chatbot- from where I have learned? What I have learned? And how can you build your first bot? Quite interesting right! Chatbot is a program that can conduct an intelligent conversation based on user's input. Since chatbot is a new thing to me also, I first searched- is there any Python library available to start with this? And like always Python has helped me this time also. There is a Python library available with name as  ChatterBot   which is nothing but a machine learning conversational dialog engine. And yes that is all I want to start my learning because I always prefer inbuilt Python library to start my learning journey and once I learn this then only I move ahead for another...

Detecting Credit Card Fraud As a Data Scientist

Another post starts with you beautiful people! Hope you have learnt something from my previous post about  machine learning classification real world problem Today we will continue our machine learning hands on journey and we will work on an interesting Credit Card Fraud Detection problem. The goal of this exercise is to anonymize credit card transactions labeled as fraudulent or genuine. For your own practice you can download the dataset from here-  Download the dataset! About the dataset:  The datasets contains transactions made by credit cards in September 2013 by european cardholders. This dataset presents transactions that occurred in two days, where we have 492 frauds out of 284,807 transactions. The dataset is highly unbalanced, the positive class (frauds) account for 0.172% of all transactions. Let's start our analysis with loading the dataset first:- As per the  official documentation -  features V1, V2, ... V28 are the principal compo...