Though statisticians are familiar with structural equation models, there have been important developments in the social science that are not widely known among biostatisticians. One example is the multiple indicators, multiple causes (model) now used widely by psychometricians to analyze data consisting of batteries of psychological tests. The Religious Orders Study, a longitudinal study of priests and nuns who agreed to autopsy at death, afforded an opportunity for my student, Alex Grigorenko, and colleagues at Hebrew Senior Life to investigate profiles of psychometric test results associated with autopsy findings of Alzheimers disease and cerebrovascular injury at autopsy.
The data consist of annual scores on a battery of 17 results of annual psychometric tests in a subset of 487 study participants for whom autopsy results are available. This talk will describe the Religious Orders Study, discuss the MIIMIC model as an important structural equation model, discuss longitudinal extensions of the MIMIC model, and then present results comparing test result profiles in three groups of study participants, those with autopsy findings of Alzheimers disease, those with evidence of cerebrovascular injury, and participants with neither of these findings.