This is a pretty basic question, but I can't find an answer by searching for different statements of the same problem.
Is there a straightforward way to test if a regression parameter is different from a non-zero value in (non-) linear regression? I can only think of one way, but it seems too roundabout. Here's an example:
INPUT PROGRAM. LOOP #I = 1 TO 10. COMPUTE X = #I. COMPUTE Y = RV.NORMAL(5,1)+x*RV.NORMAL(4,1.5). END CASE. END LOOP. END FILE. END INPUT PROGRAM. dataset name exampleData WINDOW=front. EXECUTE.
Say I want to test if the slope from linear regression deviates from 3. The only way I know how in SPSS is NLR:
MODEL PROGRAM b0=5 b1=0. COMPUTE PRED_=b0 + (3+b1)*x. NLR Y /OUTFILE='C:\temp\SPSSFNLR.TMP' /PRED PRED_ /CRITERIA SSCONVERGENCE 1E-8 PCON 1E-8.
This will give you an estimate, std. error, and 95% CI for b1, but no p value. I also get a weird feeling that this method isn't statistically "correct". Does SPSS have a built-in test for this without resorting to NLR? Is there one that gives a p value?
What if my regression model is non-linear? I suppose I can use the same approach as above, but how can I get a p value?
EDIT: I would like to clarify that I am asking about more than just SPSS methodology. Although I'd love to get an SPSS-specific answer, I am also looking for a more general statistical method to a) get a p value from regression parameters from the estimate, std. error, and degrees of freedom and b) if there's a statistical test designed to test deviation of a regression parameter from a non-zero constant.