Quantitative Lunch:Design And Data Analysis (DADA)
Linking Partial Credit Model Analyses of Longitudinal Affect Data: Comparing Anchoring Techniques
 Thursday, November 5
12:30PM - 1:45PM
Gilmer 225
Monica K Erbacher
| One of the goals of the Notre Dame Successful Aging Study is to measure and
model a variety of psychological phenomena in adults ages 60-75 over a period
of 56 days. As one aspect of this study, the present project aims to evaluate
the measurement of positive (PA) and negative affect (NA), by examining the
performance of items from the Positive and Negative Affect Schedule as well as
additional items, and to longitudinally model PA and NA. Item response models
(IRMs) facilitate the accomplishment of both goals by providing item
statistics with which to evaluate measurement and trait level scores for use
in longitudinal models. Trait level scores from two different IRMs are not
estimated on the same scale, and thus estimates must be anchored to a common
scale to use IRMs on data from separate occasions. This presentation compares
four anchoring techniques that can be used to link IRM analyses across 56 days
of measurement, in order to determine which method produces the most desirable
trait level score estimates. Results differ depending on whether the
measurement of PA or NA is being examined. When measuring PA, all four
techniques perform adequately, while for the measurement of NA, one method
performs noticeably worse than the others. Possible reasons for this
difference include missing data and differential use of response categories
across time points for NA items. |
For More Information:
Sponsored by Other
, mdh9g
Submitted by mdh9g
|