LM and generalized linear models (GLM) were used to analyze the neuropsychological performance data with age as a covariate. Log transformations were applied to time measures and analyzed using LM assuming a normal distribution and identity link function. Integer-valued variables (e.g., EM) were analyzed using zero-inflated count models combining a point mass at zero with a negative binomial distribution model to reflect the over-dispersion of the data with a log-link function using GLM. For the VST measures (e.g., Span), cumulative logit models (Agresti, 2012) were applied. The log-likelihood test was used to compare the models, and goodness of fit test was used to select the most parsimonious model.
