Page 1 of 1

[SOLVED] Factors in regression

Posted: Fri Sep 22, 2017 5:35 pm
by vmoreno
Hi,
Thanks again for this piece of programming art.

I thought that defining a variable as Nominal would treat the variable as factor (both if character or numerical) and would be used as such in linear regression, but is not the case. Predictors are only treated as continuous, not categorical.

I think treating Nominal variables as factors would be a plus and should not be too complicated.
More difficult would be to allow for fitting interactions among variables in the linear regression menu. Now we can do it pre-calculating the product terms as new variables with the new computed variable option.

Victor

Re: Factors in regression

Posted: Tue Oct 03, 2017 11:24 am
by mcfanda@gmail.com
Hi, try this new module https://github.com/mcfanda/gamlj

(it needs testing so any feedback will be highly appreciated)

Re: Factors in regression

Posted: Mon Nov 06, 2017 4:16 pm
by Ravi
Hi,
Both of these things, factors and interactions, are hopefully gonna make their way into the linear regression within a couple of weeks. I do encourage you to try out the gamlj module, it allows you to fit a bunch of linear models in a very flexible way. Good stuff.

Cheers

Re: Factors in regression

Posted: Mon Mar 05, 2018 2:55 am
by Ravi
We did a major overhaul of the linear regression in our latest version (0.8.3). It now allows for factors and interactions. Thanks for your request!