ion years, and assigned
medication. The correlation of repeated measures within each
subject was modeled with an unstructured covariance matrix. In
the analysis of CPT, omission errors and commission errors were
categorized into 4 groups, because of skewed distribution of the
numbers of errors. Between each group, we performed GEE to test
whether there is a relationship with other clinical measures. Beta
coefficient refers to degree of association between each cognitive
outcome scores and HAM-A scores, also at each assessment occasion,
same as MEM and GEE.