How to fit a nonlinear mixed model regression with parameters of the model - real numbers

I have experimental data from 4 experiments with 24 subjects divided in two age groups. I want to fit the data with a nonlinear mixed regression model. I have two difficulties: how to specify that the parameters of the model should be real positive numbers and how to specify that the parameters could depend on the age group and on the experiment.

Answer 1

[David Chorlian]

You need to describe your model in more detail. What are you trying to find out? What, if any, is the relation between the different experiments? What is the nature of your data? Once this is known, it will be possible to give meaningful answers.

[Nadejda]

The experiments differ in how we manipulated either the temporal or the spatial characteristics of the stimuli. Thus, all these are fixed factors. We assume that a common model could be applied to all data from all experiments, but the value of the parameters would depend on the age group and the experimental conditions. The experimental data are discrimination thresholds. They could be considered as a measure of variance. The model that we want to fit is the so called variance summation model i.e. the discrimination threshold th=sqrt(int_noise^2+ext_noise^2/N). We manipulated the ext_noise. The parameters of this model (int_noise, N) should be positive real numbers and most probably int_noise(experiment, age_group) and N(experiment, age_group). The age_group is within subjects factor, but I want to regard the subjects in each group as random. Each subject participated in all experiments and thus, their responses might be correlated.