Hi,

I'm trying to fit data from 10 different subjects to a nonlinear mixed effect model. The model is a sigmoid function with 4 parameters, all expected to be different from subject to subject, but whereas two of those parameters are drawn from a normal distribution and therefore can be replaced by population mean and population variability parameters (the mean and std of the normal distribution in question), the other two should be treated as subject-specific, rather than drawn from a normal distribution. That means that they should keep individual values for each subject.

Is this possible in Stata?

The function in question (for subject i) is:

Y_i(X) = Bottom_i + (Top_i-Bottom_i)/(1+10^((LogIC50_i-X)*HillSlope_i))

The LogIC50 and HillSlope parameters are normally distributed and I am interested in an estimate of their population means and variability, but the Top and Bottom parameter (defining the top and bottom of the sigmoid fit) can vary (fairly wildly) from subject to subject.

So the mixed effect model I'm trying to employ is

Y_i(X) = Bottom_i + (Top_i-Bottom_i)/(1+10^((LogIC50_mean + sigma_LogIC50*epsilon1_i-X)*(HillSlope_mean + sigma_HillSlope*epsilon2_i))

Any suggestions, short of renormalizing all data from Bottom-Top to 0-100%?

Cristian