Parametric Statistical Inference for Comparing Means and Variances

Area of Science:

• Vision science
• Biostatistics
• Ophthalmology

Background:

• Accurate statistical comparison of data groups is crucial in vision science research.
• Parametric statistical tests are commonly used but require normally distributed data.
• Understanding these tests ensures reliable and reproducible research findings.

Purpose of the Study:

• To provide visual scientists with methods for comparing two or more data groups using parametric statistical tests.
• To demonstrate approaches for analyzing normally distributed data, with non-parametric alternatives for non-normal distributions.
• To guide the interpretation of statistical software output for these analyses.

Main Methods:

• Review of parametric statistical approaches including two-sample t-test, Analysis of Variance (ANOVA), and paired t-test.
• Application of linear mixed-effects models for analyzing repeated measures data.
• Use of retinal thickness from optical coherence tomography (OCT) as an example dataset.

Main Results:

• Demonstration of analyzing differences between group means using various parametric methods.
• Discussion of follow-up procedures for pairwise mean comparisons in multi-group studies.
• Examination of the assumption of equal variances and methods for testing it.

Conclusions:

• Parametric inference tests for comparing means are outlined, with guidance on interpreting statistical software output.
• Critical assumptions of these tests and methods for verification are discussed.
• Emphasis on accounting for correlations in repeated measures data for rigorous and validated vision science research.