College of Liberal Arts
Biometrically informative datasets (e.g., datasets collected from genetically related individuals) are becoming increasingly available in the social sciences. For example, the UMN psychology department has at least three large scale twin-family projects that include biometrically informative data from many thousands of individuals (twins reared together, twins reared apart, parent of twins, etc.). These data are also routinely used in basic social science research that does not involved biometric model fitting (e.g., research aimed at elucidating the phenotypic structure of personality and vocational interests variables). Because data from genetically related individuals do not satisfy the i.i.d. (independent and identically distributed) assumptions of most statistical procedures, questions have been raised about the accuracy of bootstrap confidence interval coverage rates that are generated from such data. This computer-intensive project will use Monte Carlo simulation to provide initial insights into the consequences of treating twin-family data as singleton data. The R code for running the simulations has been written and can be run in parallel on the supercomputers.