Conditionally Specified Joint Distribution for Self-Proxy Data in Gerontology Studies (306486)Nagaraj Neerchal, University of Maryland Baltimore County
*Nadeesri Wijekoon, University of Maryland Baltimore County
Keywords: Self-Proxy data, Conditionally specified models
In longitudinal studies of health care areas, subjects may not be able to respond to questions over time due to various reasons. Therefore, researchers might have to rely on the answer of another person (proxy) who’s related to the subject. The current approach is either to treat the self and proxy data as interchangeable or simply analyze self and proxy data separately. The first approach leads to biased estimates whereas the later results in two sets of estimates without an obvious way of combining them. Thus, there is a need to develop a single framework method for analyzing both subject data and proxy data. We propose a conditionally specified joint distribution which accommodates both self and proxy data coming from different sources. We will explore the theoretical aspect of the derived conditionally specified joint distributions specially regarding parameter estimation. In addition, we will introduce “R shiny” applet which we can use to generate data from conditionally specified self-proxy joint distribution.