Find the right test for your research
Once you have data, head back to Pearl's main page to actually run the analysis.
What are you trying to learn?
Pick the goal that best describes your research question.
Are subjects the same across conditions?
In a within-subjects (paired) design, the same person is measured multiple times. In between-subjects, different people are in different groups.
How many times are subjects measured?
Pre/post is two timepoints. Baseline + multiple follow-ups is 3+.
How many predictors do you have?
If you're just looking at the link between two variables, that's one predictor.
What's your outcome (dependent variable)?
The variable you want to compare across groups.
How many outcomes (DVs) are you measuring?
Most studies have one outcome. Multiple DVs call for MANOVA/MANCOVA.
How many groups are you comparing?
Counts the number of levels of your grouping variable.
How many grouping variables (IVs)?
Two IVs gives you a factorial design (main effects + interaction).
What about your predictors / independent variables?
Type of the variable(s) you think influence the outcome.
Do you have a covariate to control for?
A nuisance variable you want to statistically remove (e.g., pre-test score, age, baseline).
Is your outcome roughly normal?
Parametric tests (t-test, ANOVA, Pearson) assume the outcome is roughly normally distributed. If your data is heavily skewed, ordinal (Likert ranks), or you have a small sample with outliers, the non-parametric alternative is safer.
How many categories does your outcome have?
Two = binary (yes/no, pass/fail). Three or more = multinomial (A/B/C).
—
Want to refine? Start over with different answers, or head back to Pearl to upload data and run the test.