topic Re: extracting parameters from proc nlmixed in SAS Procedures

extracting parameters from proc nlmixed

csetzkorn — Thu, 16 Nov 2017 15:47:37 GMT

I tend to use:

ParameterEstimates = WORK.ParameterEstimates
FitStatistics = WORK.Model_Fit
ANOVA = Work.ANOVA
SelectionSummary = WORK.SelectionSummary;

to extract model parameters and other stats when I use proc glmselect. I am trying to do something similar for proc nlmixed. Looking at this preliminary/simplified code:

ods select none;
proc nlmixed data=AnalysisData;
parms s2u1=1;
y = exp(-X);
model Y1 ~ normal(y, s2u1);
ods output
parameters = Work.Parameters;
run;
ods select All;

I obtain:

s2u1 NegLogLike
1 553224.908

Where is the parameter for X? Any input would be very much appreciated. Thanks!

Re: extracting parameters from proc nlmixed

ballardw — Thu, 16 Nov 2017 16:08:42 GMT

I think if a variable is not on the PARMS statement when PARMS is used then it is not included in the ParameterEstimates output

Re: extracting parameters from proc nlmixed

Reeza — Thu, 16 Nov 2017 16:09:29 GMT

Initial Parameters are in the Parameters tables.

Estimates are in the ParameterEstimate table.

Does that provide what you're looking for?

@csetzkorn wrote:

I tend to use:

ParameterEstimates = WORK.ParameterEstimates
FitStatistics = WORK.Model_Fit
ANOVA = Work.ANOVA
SelectionSummary = WORK.SelectionSummary;

to extract model parameters and other stats when I use proc glmselect. I am trying to do something similar for proc nlmixed. Looking at this preliminary/simplified code:
ods select none;
proc nlmixed data=AnalysisData;
parms s2u1=1;
y = exp(-X);
model Y1 ~ normal(y, s2u1);
ods output
parameters = Work.Parameters;
run;
ods select All;
I obtain:

s2u1 NegLogLike
1 553224.908

Where is the parameter for X? Any input would be very much appreciated. Thanks!

Re: extracting parameters from proc nlmixed

csetzkorn — Thu, 16 Nov 2017 18:44:32 GMT

Hi all,

Sorry actually the following code snippet works fine:

ods output
Parameters = Work.Parameters
FitStatistics = WORK.Model_Fit
;

Sorry!