From: Nathan on
Hi all,

I am trying to develop a multiple regression model for a dataset. I want to specify different functional forms for certain explanatory variables in my model, such that I could have something like the following:

Y = p1*x + exp(p2*x2) + p3*x3 + p4

where x and x3 have linear responses on Y but x2 has an exponential response. I also need to weight each data point (something that can be done if I can use the Curve-Fitting Toolbox), but I can weight the data points manually if necessary. Eventually I would also like to assess interactions between the different variables.

Can this be done using MATLAB (without programming the statistics on my own)? I have been trying to use the Curve-Fitting Toolbox, and although you can specify functional forms, this seems to only work for a single explanatory variable. I do have the Statistics Toolbox as well, but it seems to only work for all-linear multiple regressions.

Thanks!
Nathan