coefficient determination in proc genmod

[sasgiaz]

Hi, I need to get the coefficient determination value from the proc genmod.

I've read the manuals on SAS about this proc but I could not find the statement.

Is that mean this proc can not produce the value of coefficient determination?

Thanks before.

[PaigeMiller · Accepted solution]

In general, the coefficient of determination (also known as R-squared) is only meaningful for linear model, such as those fit via PROC GLM or PROC REG. Once you moved beyond linear models into generalized linear models (such as PROC GENMOD), this statistic can be problematic, and as far as I know, there is no general consensus among statistician on how to measure the fit of a generalized linear model. Here is one such discussion: https://jonlefcheck.net/2013/03/13/r2-for-linear-mixed-effects-models/ Google finds many other papers on this subject as well.

[sasgiaz]

Honestly I wanted to estimate the parameter in regression analysis with maximum likelihood estimation and I thought the proc genmod would do so. But now I think I was wrong?

[PaigeMiller]

---

@sasgiaz wrote:

Honestly I wanted to estimate the parameter in regression analysis with maximum likelihood estimation and I thought the proc genmod would do so. But now I think I was wrong?

---

GENMOD will estimate the parameter in regression analysis, using maximum likelihood estimation. In the limited case of ordinary least squares regression, where the errors are normal, maximum likelihood and least squares ought to give identical parameter estimates. Nevertheless, GENMOD does not produce a coefficient of determination.

[sasgiaz]

understood. thankyou so much!