topic Re: [SAS EM] Fourth quantile of interactive binned variable does not show up in regression formula in SAS Data Science

[SAS EM] Fourth quantile of interactive binned variable does not show up in regression formula

PearlyQuek — Thu, 22 Apr 2021 03:24:57 GMT

Hi everyone,

I had used the interactive binning node to bin my interval input variables into 4 quantiles:

However, when I connect it to the regression node, only 3 quantiles show up:

Would there be a reason why only quantiles 1, 2 and 3 show up in the parameter list?

Thank you!

Re: [SAS EM] Fourth quantile of interactive binned variable does not show up in regression formula

gcjfernandez — Thu, 22 Apr 2021 11:49:05 GMT

when performing optimal binning you are transforming interval variable to nominal variable with 4 levels. The last level by default is treated as the reference group and the first three levels are compared with the reference levels.

Re: [SAS EM] Fourth quantile of interactive binned variable does not show up in regression formula

PearlyQuek — Sun, 25 Apr 2021 09:37:40 GMT

Thank you for your help! We were having a hard time interpreting the results until we figured that the fourth quantile was absorbed into the intercept.

Re: [SAS EM] Fourth quantile of interactive binned variable does not show up in regression formula

gcjfernandez — Sun, 25 Apr 2021 17:26:34 GMT

Yes absorbing the last level with the intercept is the default option in REG node. If you fitting a Logistic regression you cold suppress the intercept and change input coding to GLM then you see all four coefficients for your binned variable.

Re: [SAS EM] Fourth quantile of interactive binned variable does not show up in regression formula

PearlyQuek — Mon, 26 Apr 2021 06:16:16 GMT

Thank you so much for your guidance!