Mixed Effects with Cox model (multiple events)
On 14-03-11 11:26 AM, Maria Helena Mourino Silva Nunes wrote:
Dear all, I want to describe the time interval between mammography screenings as a function of several clinical and demographic variables. For each woman, I have several events and each event corresponds to the time interval between two consecutive mammography screenings. Also, women's data sets come from ten different primary health centres. I have data from 1995 up to 2013. I was thinking on developing a Cox regression model with multiple events (as, for instance, Andersen and Gill model). Having in mind there are differences between the primary health centres, I was wondering if I could use a mixed model where women were nested on the primary health centres. I've already tried to use the "coxme" package but it didn't work. Is it possible to use a mixed model in this context? What package should I use? Thanks, in advanced, for your attention. Best regards, Helena Mouri?o Nunes.
For CPH you will probably have to use a specialized package. The survival package has a frailty() function that "allows one to add a simple random effects term to a Cox or survreg model" (and choose among gamma, gaussian, or t-distributed random effects). I don't know how this will fit in with your multi-event model, but that's where I would start ... Ben Bolker