Statistics

Choosing a statistical test

Rajib Biswas | 25 May 2020

Table of Contents:

Choosing the right statistical method to perform with your data is a critical decision to make. Without proper knowledge to the variables and what your target is, it is very difficult to get to the result that you want. There are several ways to decide what statistical test should be done with the data you have and what you want to see. I’m going to give some of the most useful examples of them here in this post.

From Intuitive Biostatistics

In the book Intuitive Biostatistics by Harvey Motulsky (Copyright © 1995 by Oxford University Press Inc.) this following beautiful table has been crafted in chapter 37 which I think is a very good starting point to decision making.

 Type of Data
GoalMeasurement (from Gaussian Population)Rank, Score, or Measurement (from Non- Gaussian Population)Binomial
(Two Possible Outcomes)
Survival Time
Describe one groupMean, SDMedian, interquartile rangeProportionKaplan Meier survival curve
Compare one group to a hypothetical valueOne-sample ttestWilcoxon testChi-square
or
Binomial test **
 
Compare two unpaired groupsUnpaired t testMann-Whitney testFisher’s test
(chi-square for large samples)
Log-rank test or Mantel-Haenszel*
Compare two paired groupsPaired t testWilcoxon testMcNemar’s testConditional proportional hazards regression*
Compare three or more unmatched groupsOne-way ANOVAKruskal-Wallis testChi-square testCox proportional hazard regression**
Compare three or more matched groupsRepeated-measures ANOVAFriedman testCochrane Q**Conditional proportional hazards regression**
Quantify association between two variablesPearson correlationSpearman correlationContingency coefficients** 
Predict value from another measured variableSimple linear regression
or
Nonlinear regression
Nonparametric regression**Simple logistic regression*Cox proportional hazard regression*
Predict value from several measured or binomial variablesMultiple linear regression*
or
Multiple nonlinear regression**
 Multiple logistic regression*Cox proportional hazard regression*

 

For further reading, have a look at this post by GraphPad.

Share this article!