Introduction

Here you can find information regarding the R-squared computed for different models and their associated inferential tests.

GLM (linear model)

The R-squared produced by GAMLj is the standard R-squared produced by virtually any statistical software . It is computed as \(SS_{model}\) divided by \(SS_{total}\). In GAMLj, it is extracted from the R model object by performance::r2(model) (Lüdecke (2020)). The adjusted R squared is computed as

\[R_{adj}^2={{SS_{model}-SS_{res} \cdot ({df_{model}/df_{res}) \cdot }}\over{ SS_{model}+SS_{res}(df_{res}+1)/df_{res}}}\]

The F-tests associated to the R-squares tests the null-hypothesis that the model predictors explain no variance, which is equivalent to compare the model at hand with an intercept-only model.

GzLM (Generalized linear Models)

For the generalized linear model, the R-squared