How to calculate the AIC, BIC, RMSE, and R2 for a Nonlinear regression model?

[ukhan]

I am using the von Bertalanffy growth function (VBGF), but the PROC STATEMENT does not provide model fitness parameters such as AIC, BIC, RMSE, and R2? Need help to revise the PROC STATEMENT.

FILENAME REFFILE '/home/u58890255/Growth  functions/sample_data.xlsx';
PROC IMPORT DATAFILE=REFFILE
	DBMS=XLSX
	OUT=WORK.IMPORT;
	GETNAMES=YES;
RUN;
PROC CONTENTS DATA=WORK.IMPORT; RUN;

title 'von Bertalanffy growth function';
data exp;
set WORK.IMPORT;
if expno=1;
run;
proc nlin data=exp method=marquardt;
parms Linf=190 k=0.05 t0=-1.5;
model sl=Linf*(1-exp(-k*(age-t0)));
output out=expp p=pLinf r=Linf_residual;
run;

Using the following PROC STATEMENT,  I can calculate the adjusted R2 for a Nonlinear regression model, but this time the following PROC also does not provide any result

proc summary data=expp;
     var Linf_residual Linf;
     output out=stats css(Linf)=sstot uss(Linf_residual)=ssres N=N;
run;
data expp;
     set stats;
     rsquared=1-(ssres/sstot);
     adjrsquared = 1-(1-rsquared)*(N-1) / (N- 3  -1);
run;
proc print data=expp;
run; 

Kind regards,

[sbxkoenk]

Success! We moved this post to "statistical_procedures" board.

[StatDave · Accepted solution]

Use PROC NLMIXED instead of PROC NLIN - it's a bit more flexible (particularly if the response is not normal) and provides AIC and BIC. For an R-square, you can use the generalized R-square discussed in this note. 

proc nlmixed data=import;
where expno=1;
m=Linf*(1-exp(-k*(age-t0)));
model sl~normal(m,s);
predict m out=out;
run;
data out; set out;
res=sl-pred;
run;
proc sql; 
select 1-(uss(res)/css(sl)) as R2 from out;
quit;

[ukhan]

Great, it works perfectly. Kind of you @StatDave. Could you please also explain what is the new parameter “s” (please check the attached screenshot)?

[Ksharp]

According to code " sl~normal(m,s) " ==> s is the std of normal .

[ukhan]

Thank you so much for the help 🙂