topic Re: How to use PROC NLMIXED with two levels of a fixed factor in Statistical Procedures

How to use PROC NLMIXED with two levels of a fixed factor

Tessie — Fri, 24 Apr 2020 19:03:20 GMT

Hi, I am trying to fit my uptake data to different models using PROC NLMIXED. However, I am a new user of NLMIXED and looking for a code that will help in differentiating the parameters of the two levels of fixed effect.

I have two plants (Cabbage and grass) and will like to be able to estimate if parameter A for cabbage is different from parameter A of switchgrass. I have attached the code and will like to know if I am on the right track.

Thanks

title 'Logistic growth model  individual parameters for plants';
proc nlmixed data=uptake;
   parms A1=0.2 to 0.5 by 0.1 A2=0.2 to 0.5 by 0.1 B1=8 to 21 by 0.5 B2=8 to 21 by 0.5   k1=0.3 k2=0.3s2e=0.01528 s2u=0.01528;
   if Plant='Cabbage' then eta=(A1/(1+exp(-k1*Day+B1*k1)));
   else eta=(A2/(1+exp(-k2*Day+B2*k2)));
    p = eta;
   model Amount ~ normal(p,s2e);
   Estimate 'A1-A2' A1-A2;
    Estimate 'B1-B2' B1-B2;
   Estimate 'k1-k2' k1-k2;
   predict p out=uptakeout;
  run;

Re: How to use PROC NLMIXED with two levels of a fixed factor

SteveDenham — Mon, 27 Apr 2020 14:00:08 GMT

Your code looks exactly like that found in the SAS/STAT 14.1 documentation for Example 82.4, so if you are comfortable with the model you have chosen to fit, I would say go ahead. If you have execution time problems, consider rewriting the exponentiated term in the denominator as exp( -k * (Day - B)) with the proper subscripts for k and B. One addition and one multiplication should run faster than one addition and two multiplications.

SteveDenham

Re: How to use PROC NLMIXED with two levels of a fixed factor

Tessie — Tue, 28 Apr 2020 00:12:37 GMT

Thank you SteveDenham.