topic Re: Proc NLIN and adjusted R² in Statistical Procedures

Proc NLIN and adjusted R²

soumri — Fri, 08 Jul 2016 11:09:03 GMT

Hello everyone, I used a Proc NLIN to fit a nonlinear function, but as this regression has no intercept, I have no R². I think, in this case, the adjusted R² is used [R²a = 1-SS (residuel) / SS (corrected total)], (mentioned in several publications). is anybody can help me to calculate it.

Re: Proc NLIN and adjusted R²

PGStats — Fri, 08 Jul 2016 18:59:27 GMT

Adjusted R-square is adjusted for the number of parameters fitted, not for the absence of an intercept. Removing the intercept changes the definition of R-square by adding the assumption that in the absence of independent variable effects, the response would be zero (instead of some mean value). Although removing the intercept is sometimes convenient to get some estimates, it is not compatible with estimating the proportion of variance explained by your model.

Re: Proc NLIN and adjusted R²

soumri — Mon, 11 Jul 2016 09:01:54 GMT

@PGStats

Thank you very much PGstats for wealthy enough information that you have provided me, in addition, there is a fairly long conversation entitled "Adjusted R^2 (adjusted R square)" which contains the answer to my question with further details.