How do I choose the right statistical test for my data?

Choosing the Right Statistical Test

With so many statistical tests available, it is easy to feel lost. The good news is that the choice usually comes down to a few questions about your research question and your data. In this blog, we'll walk through those questions and match common situations to common tests.

Question 1: What Is Your Aim?:

  • Are you comparing groups, for example men and women, or three teaching methods?
  • Are you looking at a relationship between variables, for example workload and stress?
  • Are you predicting an outcome from one or more variables?

Question 2: What Type of Data Do You Have?:

  • Categorical (nominal) data: groups or categories, such as gender or department.
  • Ordinal data: ranked categories, such as education level.
  • Continuous (scale) data: numerical measurements, such as age, income or a scale score.
  • Identify both your outcome (dependent) variable and your predictor or grouping (independent) variable.

Question 3: How Many Groups, and Are They Related?:

  • Two groups or more than two?
  • Independent groups, where different people are in each group, or related groups, where the same people are measured twice?

Question 4: Are the Assumptions Met?:

  • Many common tests, called parametric tests, assume roughly normal distributions and similar variances across groups.
  • If these assumptions are not met, a non-parametric alternative may be more appropriate.

Common Matches:

  • Comparing two independent groups on a continuous outcome: independent-samples t-test (or Mann-Whitney U).
  • Comparing the same people at two times: paired-samples t-test (or Wilcoxon signed-rank test).
  • Comparing three or more independent groups: one-way ANOVA (or Kruskal-Wallis test).
  • Testing the association between two categorical variables: chi-square test of independence.
  • Measuring the relationship between two continuous variables: Pearson correlation (or Spearman correlation).
  • Predicting a continuous outcome from several variables: multiple linear regression.
  • Predicting a two-category outcome, such as yes or no: binary logistic regression.

Choosing a statistical test is a logical process. Clarify your aim, identify your variable types, count your groups, check your assumptions, and then select the test that fits. Plan this before you collect data, and discuss your choices with your supervisor or a statistics adviser.