How to find best polynomial linear regression model (equation) for multiple datasets?

Hello,

I have multiple data sets, each with 3 independent variables and 1 response. i.e., [y, x1,x2,x3].

I would like to find the best functional form (linear regression) that would work best for ALL of the data sets. The coefficients will be different for different data sets.

I tried stepwiselm individually for each data set and obviously get a different equation and different number of terms.

What I am looking for is stepwiselm that works on multiple datasets simultaneously.

Thank you for your help.

-V

[Jeff Miller]

Do you want the same coefficient values in the models for all data sets, or do you just want the same linear and polynomial terms with the independent variables--but possibly different coefficients?

[vaatum]

Thank you. I just want the same linear and polynomial terms with the independent variables. The coefficient values will be different for each data set.

Thanks.

[Jeff Miller]

Too bad--that is much harder.

I doubt you can find a "stepwiselm that works on multiple datasets simultaneously". If not, your only option (as far as I know) will be to fit all the data together within one combined dataset. You will have to set up "dummy variables" to code the datasets and also to allow for changes in coefficients across datasets. Then you will probably have to guide the stepwise model-fitting process yourself rather than having the search done automatically.