[R] Accounting for within family correlation in genetic analysis

John Sorkin jsorkin at grecc.umaryland.edu
Thu Dec 8 04:10:30 CET 2005

I am hoping for help with a genetic analysis.

I am trying to perform an analysis of the relation between genes at a given locus (rs2304795) and a phenotypic trait (zerotg). Multiple subjects are recruited from each family (and so share a part of their genome and are correlated). Family groups are identified by the variable FAMILY. For each subject multiple measurements are made of the trait of interest (zerotg) over the course of several hours (time). The data within each subject are of course correlated and over time each subject's response is curvilinear. I know how to deal with the within subject correlation using a growth-curve analysis:

	zerotgQuad0ML.lme <- 
		lme (zerotg ~ time + time^2 + rs2304795 + rs2304795 * time +   rs2304795 * time^2, 
			data = GC, random =  ~ 1 + time + time^2 | HAPI, 
                      		na.action = na.omit, method = "REML") 

but I don't know how I can modify the model to account for the within-family correlation. 

I would appreciate any suggestions for modifications to the code above that would allow me to account for the within-family correlation of observations.


John Sorkin M.D., Ph.D.
Chief, Biostatistics and Informatics
Baltimore VA Medical Center GRECC and
University of Maryland School of Medicine Claude Pepper OAIC

University of Maryland School of Medicine
Division of Gerontology
Baltimore VA Medical Center
10 North Greene Street
Baltimore, MD 21201-1524

jsorkin at grecc.umaryland.edu

More information about the R-help mailing list