topic Re: Change the default -1 1 coding of categorical variable to 0 1 coding in Discussions

Change the default -1 1 coding of categorical variable to 0 1 coding

Jogger560 — Thu, 12 Sep 2024 17:41:44 GMT

Dear JMP community,

I found that when fitting a multiple linear regression (y ~ X1 + X2, X1 is a continuous and X2 is a categorical variable), by default JMP defines the dummy variable X2 as -1 1. Is there a way to change it to 0 1, which will be the same as the default dummy variable coding as R? Thanks!

Re: Change the default -1 1 coding of categorical variable to 0 1 coding

MRB3855 — Mon, 16 Sep 2024 08:23:13 GMT

Hi @Jogger560 : In the red triangle pull down menu at the top of the Fit Model output, there is an option for Indicator Parameterization Estimates.

Re: Change the default -1 1 coding of categorical variable to 0 1 coding

statman — Mon, 16 Sep 2024 13:46:43 GMT

Advice for coding of factor levels consistently is equivalent-distant, centered on 0. So -1, 1 for 2-levels, -1,0,1 for 3-levels.

This a quote from Mark Bailey: Coding benefits the analysis in several ways:

The parameter estimates are comparable.They are independent of the associated scale (units).
The parameter estimates are interpretable. The intercept when estimated with real values is the response when all your factors are zero. What does that mean? With coding, the intercept is the mean response at the origin (center) of your design. The other parameters still represent the change in the response for 1 unit change in the factor, but in the coded space, 1 unit change is half the range. Double the estimate and that is the change in the response over the full range of the factor.
The parameter estimates are uncorrelated or minimally correlated. This preserves power for the significance tests.

Re: Change the default -1 1 coding of categorical variable to 0 1 coding

MRB3855 — Mon, 16 Sep 2024 13:53:56 GMT

Hi @statman : I understood his question to be about the categorial factor (X2 in his OP); you seem to be discussing the covariate X1?

Re: Change the default -1 1 coding of categorical variable to 0 1 coding

statman — Mon, 16 Sep 2024 17:44:48 GMT

The coding applies to any type of factor in an experiment, continuous or categorical. This allows for easy comparison of the coefficients for analysis.

Re: Change the default -1 1 coding of categorical variable to 0 1 coding

MRB3855 — Mon, 16 Sep 2024 14:15:24 GMT

Hi @statman : Perhaps, but your first and third bullets don't seem relevant for X2 categorical.

Re: Change the default -1 1 coding of categorical variable to 0 1 coding

statman — Mon, 16 Sep 2024 16:33:17 GMT

I just copied and pasted Mark Baileys comments on factor coding...I didn't want to cherry pick what he wrote:

https://community.jmp.com/t5/Discussions/DOE-Code-and-Uncoded-levels-of-factor/td-p/61864

Re: Change the default -1 1 coding of categorical variable to 0 1 coding

Jogger560 — Mon, 16 Sep 2024 16:52:21 GMT

Thanks very much! I also need to get the correlation of estimates under the 0, 1 coding coding, is there a way to do that?

Re: Change the default -1 1 coding of categorical variable to 0 1 coding

MRB3855 — Tue, 17 Sep 2024 08:03:11 GMT

Hi @statman : Indeed, I don't disagree with what Mark wrote wrt continuous variables. But, that is not relevant for categorical factors.

https://www.jmp.com/support/help/en/18.0/index.shtml#page/jmp/nominal-factors.shtml#

Re: Change the default -1 1 coding of categorical variable to 0 1 coding

MRB3855 — Tue, 17 Sep 2024 08:24:07 GMT

Deleted

Re: Change the default -1 1 coding of categorical variable to 0 1 coding

MRB3855 — Tue, 17 Sep 2024 08:37:04 GMT

Hi @Jogger560 : Hmm, Some JMP folks may want to jump in here but I don't think that is directly available. You can get it yourself though. The Var-Cov matrix of the OLS Estimates is defined on page 8 here.

https://web.stanford.edu/~mrosenfe/soc_meth_proj3/matrix_OLS_NYU_notes.pdf