Room: 16, Upper
Level 6E (CC)
Recent developments in measurement
theory can provide new insights into substantive issues in applied research
settings. The Measurement, Statistics and Research Design interest group
symposium includes five presentations concerning the application of such
techniques in topical areas related to psychological, sociological and
clinical aspects of gerontology. The symposium is notable for the
participation of students in two papers. Burant and colleagues present an
analysis of factor pattern differences in a Quality of Pain scale between
White and African-American patients with hip/knee osteoarthritis. Jones and
Cazorla-Lancaster also present an analysis aimed at elucidating measurement
noninvariance, and use the MIMIC (Multiple Indicators, Multiple Causes) model
approach to addressing within-Hispanic/Latino(a) group heterogeneity in
depressive symptom expression. Mast extends the MIMIC model to second order
factors, with regard to how cerebrovascular disease relates to the expression
of depressive symptoms and cognitive impairment. An important factor in
considering cognitive impairment is educational attainment, and here Glymour
and Berkman present an instrumental variables approach to untangling
socio-political heterogeneity in the meaning of educational attainment due to
variability in state education policies in inferring causal relationships
between education and later life cognitive performance. Finally, Newsom and
Rook extend the measurement theme longitudinally in an examination of social
exchanges, using a state-trait error model approach. The techniques and
modeling strategies demonstrated by the symposium participants are applicable
to many different substantive areas in gerontological research, particular in
the exploration of mechanisms underlying health disparities. |
|
1:45 PM - 3:15 PM |
||||
|
|
An
Application of 2nd Order MIMIC Models for Geropsychological Assessment
Research |
16 |
Mast |
|
|
|
Using
Natural Experiments to Estimate Education Effects on Cognitive Function |
16 |
Glymour |
|
|
|
16 |
Jones |
|
|
|
|
Trait-State-Error
Measurement Models of Stable and Changing Aspects of Social Exchanges |
16 |
Newsom |
|
|
|
16 |
Burant |
|
|
|
Room: 10, Marriott
Hall 5 (M)
This paper is motivated by the need to understand the natural history of disability in older women at both the population and individual levels. Ordinary random effects growth curve models ignore informative dropouts, which could be of great clinical importance in understanding monumental changes in functional trajectory. Our goal is to describe functional changes while accounting for informative dropouts using a two-stage approach. First, a pattern-mixture growth curve model is used to compare patterns of individual trajectories by dropout time. Second, separate random-effects models are fitted for assessing differences in mean trajectories leading to various clinically meaningful dropout events such as death. This study shows that those who dropped out earlier in the study experienced more rapid functional decline, as did those who died. Understanding such distinctive patterns of change will help identify those at risk for catastrophic declines and design interventions to maintain good function. |