I performed a Poisson regression with robust variance estimation. I have 3 independent variables. The X1 = 1(1)100. For each value of X1 there are 500 to 2000 observations. The X2 is a deterministic, non-linear function of X1, and so is the X3. In other words, for each value of X1, all observations of X2 are the same. For each value of X1, all observations of X3 are the same. I want to use the margins function wherein for each value of X1 (i.e., at(X1=(1(1)100)), the mean of X2 is used and the mean of X3 is used. I do not want to use the mean of X2 among all values of X1. Is there a way to make this specification?