Statistical Tests

#Math
How to choose a statistical test

Broader Categories

A way to determine if groups are different from each other. More precisely, it's a way to test whether measurements between two or more groups indicate that their population means are significantly different

An n-way or n-factor statistical test compares n independent variables

Compare Means

Assume variances between groups are equal

2 Groups (e.g. Control vs. Drug)

Same Subjects Measured Twice (Before and After)

Pairwise Differences Normal

(i.e. )

Pairwise Differences Non-Normal

  • Wilcoxon Matched Pairs Signed Rank Test (non-parametric)

Each Subject Measured Once

Each Group Normal and Variances Equal

Each Group Normal But Variances Not Equal

One or Both Groups Non-Normal

>2 Groups (e.g. Drug A vs. Drug B vs Drug C)

One Factor (Drug)

Each Subject Measured More Than Once

Pass Normality
Assume Sphericity
Don't Assume Sphericity
Reject Normality

Each Subject Measured Once

Pass Normality and Variances Equal
Pass Normality But Variances Not Equal
Pass Normality and Variances Equal

Two Factors (Drug, Diet)

  • Same as One Factor except if there's no normality (so you can't use a parametric method), then you have to consider a transformation. I.e. there's no non-parametric methods

Three Factors (Drug, Diet, Time)

  • Same as Two Factors

Two Or More Groups

Discrete Dependent Variable

Relationship between X and Y

Correlation (Strength of Relationship)

Linear Regression (Slope)

Multiple Regression

Nonlinear Regression

Categorical Data

Survival Analysis