Choosing the Right Statistical Test

Over the past 10 years across my roles as an undergraduate researcher, graduate researcher, quantitative analyst, Ed Tech researcher, and data scientist, I have frequently needed to use statistics to answer important questions. Usually my experience goes something like this:

“Have I run a statistical test like this before?”


“What is the nature of my data?”

“Categorical data, multiple groups, relationship.”

(me typing in Google) “What statistical test should I use to examine the relationship between multiple categorical variables?”

(15 minutes later)

(me still reading a few articles on which test I should run)

“I should use Cramer’s V test!”

“Now what are the assumptions for a Cramer’s V Test, when can I reliably use this test, and how do I run the analysis?”

(15-20 more minutes on Google and Stack Exchange)

“Nice. Now how do I interpret the results?”

This process is painful, time consuming, and sometimes results in running the wrong test due to a bad source or interpreting a source incorrectly.

To solve this problem, I developed StatsTest is an easy to use workflow tool that guides you to the correct statistical test based on your research question and data.

StatsTest is designed to be accessible to anyone with data but without statistical expertise. This means the site is free of statistical jargon and includes high-quality images to explain complicated concepts.

Figuring out which statistical test is the right test to use in StatsTest takes 1 minute, compared to the 30-60 minutes of reading Google, Stack Exchange, and Research Gate to try to find the right test, understand the assumptions, and figure out how to run the test.

Since starting development on this tool, I have used it twice in my own work to improve my productivity and do things in a faster and easier way. Once I needed to figure out which statistical test to run, and once I wanted a quick estimate for the sample size needed based on estimated effect size (these tables are included in each methods page).

Hopefully will be useful to someone else with data but without extensive statistical expertise.

StatsTest is currently under active development, meaning not all of the methods are fully developed. Below are some of our finished methods if you want to take a look.

Single Sample T-Test

Single Sample Z-Test

Single Sample Wilcoxon Signed-Rank Test

Independent Samples T-Test

Independent Samples Z-Test

Mann-Whitney U Test

I’d love to hear if this would be useful to you or not. Let me know if you have ideas or suggestions for me.

Leave a Comment

Your email address will not be published. Required fields are marked *