topic Re: GENMOD in Statistical Procedures

GENMOD

R_mont — Tue, 03 Mar 2020 13:57:15 GMT

Good day, I am quite new with statistics and I have been working with a linear regression model using PORC GENMOD (with different link options, since not all my variables are normally distributed).

I am wondering which is the effect parameter that we interprete in this regression (the one from the Maximun likelihood parameter estimates table or the one with the differences of the response variable Least squares means)???

and if it should be interpreted differently?

Thank you a million!

Rebeca

Re: GENMOD

PaigeMiller — Tue, 03 Mar 2020 14:22:03 GMT

Good day, I am quite new with statistics and I have been working with a linear regression model using PORC GENMOD (with different link options, since not all my variables are normally distributed).

Link options are appropriate when the response variables is not normally distributed. It is wrong to pick link options because the predictor variables are not normally distributed.

I am wondering which is the effect parameter that we interprete in this regression (the one from the Maximun likelihood parameter estimates table or the one with the differences of the response variable Least squares means)???

The coefficient (or effect) of a variable is shown under "Estimate", as in this example: https://documentation.sas.com/?cdcId=pgmsascdc&cdcVersion=9.4_3.4&docsetId=statug&docsetTarget=statug_genmod_examples01.htm&locale=en

Re: GENMOD

R_mont — Fri, 06 Mar 2020 08:52:07 GMT

Thank you Paige Miller! Yes the response variables are the ones not normally distributed.

Thanks a lot for the example 🙂

Re: GENMOD

R_mont — Fri, 06 Mar 2020 13:42:29 GMT

Sorry, I have another question, how do I back transformed the values to its original skala? I thought by using the option exp but it only works for log trnasformed variables and not for all the ones that were transformed to an specific lambda for example: link=power(-0.21)

🙂

Thanks a lot