ATLASResearch
methods
Quantitative/ Analysis

t-tests

Compare means with the right dependence model

A t-test estimates and tests a mean against a reference or another mean. Independent-group Welch tests allow unequal variances; pooled Student tests assume equal variances. Paired tests analyse within-pair differences, while one-sample tests use one set of observations. The unit of independence and chosen estimand determine the appropriate version.

WHEN IT FITS

Use for a prespecified comparison involving a continuous outcome and one mean or two conditions. Establish whether observations are independent or paired before selecting the test and uncertainty estimate.

Strengths

  • Direct estimate of an interpretable mean difference
  • Welch handles unequal group variances without preliminary switching

Limitations

  • Outliers and severe skewness can impair small-sample inference
  • Repeated unplanned tests inflate false-positive risk

Know the boundary

A nonsignificant t-test does not demonstrate equivalence; use an equivalence design with justified bounds for that claim.

USED ACROSS
Banking & financeBusiness & MBAPsychologyComputer scienceEducation