COMPARING COGNITIVE BATTERIES: UNIFORM DATA SET 2 (UDS-2) AND UNIFORM DATA SET 3 (UDS-3) IN NON-HISPANIC WHITE PARTICIPANTS AND THOSE FROM UNDERREPRESENTED GROUPS ENROLLED IN THE WISCONSIN ALZHEIMER'S DISEASE RESEARCH CENTER'S CLINICAL CORE

Abstract

The network of 32 Alzheimer's Disease Centers employs a battery of neuropsychological measures. The cognitive data collected with this battery are part of the larger Uniform Dataset (UDS). In 2015 the UDS battery was updated from UDS-2 to UDS-3. Crosswalk studies have compared tests included in UDS-3 to those from the UDS-2. However, it is not yet clear if the two batteries demonstrate similar overall structure. Nor has the factor structure for subgroups been compared. The goals of the present project were to examine the factor structures of the UDS-2 and UDS-3 cognitive batteries, and to compare the underlying factor structure of data obtained from participants from underrepresented groups (URG) to that from non-Hispanic whites. Principal component analyses (PCA) were done using the singular value decomposition method in R (V 3.3.0), using variance-scaled data from cognitively healthy individuals sourced from the UDS-2 and UDS-3 tests. PCA were first performed with the full sample and secondly comparing non-Hispanic Whites to URG participants. The principal component analysis (PCA) of the full sample revealed one main component of cognition accounting for 41% of the variance in Uniform Data Set 2 and 35% of the variance in Uniform Data Set 3. Tables 1 and 2 display the PCA results for UDS-2 and UDS-3. For the UDS-2 PCA, all 3 additional components accounted for an additional 31% of the variance. An additional 28% variance was accounted for with the remaining components described in the UDS-3 PCA. Comparing non-Hispanic whites and URG participants, the PCA results were similar for both the UDS-2 and UDS-3. The cognitive batteries included in the Uniform Data Set 2 and Uniform Data Set 3 each produced a similar factor structure, and appear to perform similarly in URG participants and non-Hispanic whites. Given the interesting and somewhat related factor structures found using principle component analyses of the two batteries, a canonical correlation analysis between the UDS-2 and UDS-3 would be of interest and is warranted given the overall structure of the data.