Hi I have a dataset including 885 observations. We are looking at the association between a set of predictors and the chance of pregnancy after embryo transfer. The response variable is dichotomous (pregnant or nonpregnant). Predictors are dichotomous (number of embryos; 1 or 2, time; before a certain date and after, embryo characteristics; A or B and two other parameters) and also size category with three levels (smaller, equal, and larger than x). A preliminary chi-square test on each of these variables shows a strong association between pregnancy and two of the predictors; recipients of two embryos and those with shape A had higher pregnancy compared to recipients of 1 embryo and those with shape B.
My question is: I am trying to run a GzLM to include all predictors and estimate the ORs. The issue that arises is that observations related to double embryo transfer (n=96) do not allow evaluation of embryo type or size at the same time, since it is two embryos that we are looking at (could be two A or 1 A and 1 B or other combinations). In other words, it makes sense to say embryo type or embryo size should be evaluated in those cases that received a single embryo. I would appreciate your advice on the right approach here. I preferably want to run all parameters in one analysis. Does it make sense if I run all predictors together and consider size and type as missing values in the case of double embryo recipients (98 missing values for 98 double embryos)?
