Simple Guide for Selecting Statistical Tests When Comparing Groups

Selecting the right statistical test can prove to be a daunting task for anyone. This infographic presents a step by step approach for the test selection process. This way of looking at various conditions to pick the appropriate tests will allow the audience to visualize and remember the process easily.

However, it is also very important to have the basic understanding of statistics, related terms and concepts. It will not be a wrong statement to make that the correct statistical selection process leads to correct decisions and inferences.

11 members like this

Comment

You need to be a member of Data Science Central to add comments!

Join Data Science Central

Comment by Sunil Kappal on November 24, 2017 at 2:47am: Sure, just to make things clear and to put it in the simplistic manner, I would say

Z Test can be used to test hypotheses about mean when the population standard deviation is known and population distribution is normal or sample size is large

Where as a T-Test can be used to test hypotheses about mean when the population standard deviation is unknown. Technically, requires population distributions to be normal, but is robust with departures from normality
Important: Sample size can be small

Comment by Bao-Anh Dang Trong on November 23, 2017 at 5:11pm: For the one-sample, we can also use z-test if the variance is known, right?

Comment by Sunil Kappal on January 30, 2017 at 5:26am

Thanks for the comment!

Sure, feel free to use whatever technique you want to pick and has worked per your experience. However, I would suggest you to keep in mind that the distinction in between a paired and a non-paired sample is more than the number of samples that one would like to test in one attempt. If you end up missing on these distinctions it can potentially turn your hypothesis testing into a never ending rigmarole

In a paired sample the data is collected for a same subject (medicines), machinery (manufacturing), same Contact Center (Service industry) etc. at two different points with two measurements that acts as before and after.
In unpaired sample the data is collected from two different and independent sources and the size of the sample may or may not be equal with an assumption that the data is gathered from normal distribution and the standard deviation is the same for both the samples.

Comment by Damaraju Pavan Kumar on January 27, 2017 at 11:35pm: I may not agree this as ANOVA can perform on non-paired samples also...

The difference between paired and non-paired samples, is the number of samples one would like to test in a single attempt.

It samples are <=2, you employ t-test and if sampls are greater than 2, you will go with ANOVA one way or two way depending on the circumstances....

RSS

Welcome to
Data Science Central

Sign Up
or Sign In