Proc nlin - test if coefficient differs from one

[rj_missionbeach]

I'm estimating a model and want to test if one of the parameters differs from one (rather than zero). That is, I'm trying to find the equivalent of a "test" statement in a linear regression. Specifically, I'd like to test if the lambda coefficient differs from one. Thanks for any help.

proc nlin data=outpn1_std;

parameters lambda=1 a=0 b=0;

model &depvar.=a+b*(

((lambda**1)/(lambda**1+lambda**2+lambda**3+lambda**4+lambda**5)*xn3)

+((lambda**2)/(lambda**1+lambda**2+lambda**3+lambda**4+lambda**5)*xn6)

+((lambda**3)/(lambda**1+lambda**2+lambda**3+lambda**4+lambda**5)*xn9)

+((lambda**4)/(lambda**1+lambda**2+lambda**3+lambda**4+lambda**5)*xn12)

+((lambda**5)/(lambda**1+lambda**2+lambda**3+lambda**4+lambda**5)*xn15) );

[SteveDenham · Accepted solution]

You can get a test of whether lambda=1 by using an ESTIMATE statement (SAS/STAT 15.1 or later). For example,

ESTIMATE 'label' lambda-1;

This should give output that looks like this:

In this case, I labeled the ESTIMATE 'community' and added this statement to the first example in the PROC NLIN documentation:

estimate 'community' alpha-1;

The approx Pr >|t| is a two-tailed test of whether alpha differs from 1, by constructing the additional estimate alpha-1, which NLIN tests against zero.

SteveDenham

[rj_missionbeach]

Perfect. Thank you Steve!