|
Activity Number:
|
342
- Measurement Challenges and Innovations in Cognitive Aging and Dementia Research: Progress from the MEthods for LOngitudinal Studies of DEMentia (MELODEM) Initiative
|
|
Type:
|
Topic-Contributed
|
|
Date/Time:
|
Thursday, August 12, 2021 : 10:00 AM to 11:50 AM
|
|
Sponsor:
|
Section on Statistics in Epidemiology
|
|
Abstract #317642
|
|
|
Title:
|
Co-Calibration of Cognitive Functioning Domains Across Different Countries When There Are No Common Linking Items
|
|
Author(s):
|
Alden Gross* and Emma L Nichols
|
|
Companies:
|
Johns Hopkins Bloomberg School of Public Health and Johns Hopkins Bloomberg School of Public Health
|
|
Keywords:
|
item response theory;
psychometrics;
cross-national harmonization;
cognitive aging
|
|
Abstract:
|
Modern psychometric methods for cocalibration rely on items in common across studies to derive a common cognitive metric. Such methods are infeasible when there are no overlapping test items, a situation in cross-national research in cognitive aging due to changes necessary to accommodate cultural and linguistic differences. We propose as a solution an approach that entails (1) identifying a common test known to be highly correlated with the cognitive domain of interest in both studies; (2) separately in each study, derive scores via IRT for the cognitive domain of interest; (3) create inverse probability weights (IPW) for membership in one study based on equated scores from the common test in step 1; and (4) using equipercentile equating to equate the domain scores in step 2 to the distribution of the metric in step 3, weighting with IPW. We applied this approach to co-calibrate executive functioning across Harmonized Cognitive Assessment Protocols (HCAPs) in the US and England. Using this method, the IRT-cocalibrated IRT executive score (gold standard) was correlated with the equipercentile-equated executive score at r=0.80. We discuss limitations and future directions.
|
Authors who are presenting talks have a * after their name.
