Understanding Generalizability Analysis - G-String-Legacy/GS_MV GitHub Wiki

'Generalizability Analysis' is a form of 'Analysis of Variance' that can be employed to determine the reliability of a task performance test. Its purpose is to provide an estimate for how repeatably a test ranks the performance of test takers under comparable circumstances. The emphasis of the following discussion is on plausibility rather than on formal stringency.

It is important to understand the terms 'task', 'score', and 'rank' as they are used in this context, as well as the statistical terms mean, and variance. The mean is a measure of a data set's communality, the variance a measure of its divergence.

G_String performs Generalizability Analysis in four stages:

1. It leads users through the data entry process step-by-step

2. It estimates the variance component associated with each design factor

3. It calculates the 'G Coefficients'

4. It permits D Studies

---

Footnote:

In this discussion we focus on the use of Generalizability Analysis in performance assessment, although it has many other applications.