I would like to do a gene x environment interaction analysis in a matched (1-1) case control samples. I referred all related previous publications and in most of the papers authors used either STATA or SAS. I got few references for performing conditional logisitic regression in R, for example using survival (clogit) package. But I couldn't find any reference for adding interaction terms in conditional logistic models in R. Can someone help me with references for interaction analysis using conditional logistic regression in R?
$\begingroup$
$\endgroup$
Add a comment
|
2 Answers
$\begingroup$
$\endgroup$
As stated here (http://www.ats.ucla.edu/stat/stata/library/sg124.pdf), interaction or effect modification...is performed by including and evaluating the significance of second or higher order terms involving the two or more variables that are postulated to possibly modify their respective effects.
I´m trying something similar with clogit() in R but have not found much info about it in the web, except this link (https://stackoverflow.com/questions/20977401/coxph-x-matrix-deemed-to-be-singular) talking about the problems/errors found when using interaction terms with the function coxph(). Since clogit() is a wrap-up of coxph(), I thought this could be useful.
$\begingroup$
$\endgroup$
To create an interaction effect between two variable (X1, X2). Simply take the products of the variables (X3 = X1 * X2) and then add this new variable in your model in addition to the 2 other ones => Y = X1 + X2 + X3. Depending on the type of variables (i.e., discrete/binary vs. continuous) you might want first to mean center (X1, X2).
