Question:
Is there a systax to do bootstrapping for OLS estimators in SPSS?
we know some softwares like TSP could do this job.
i would really appreciate someone's help in this regard.
Answer:
The CNLR procedure for constrained nonlinear regression does unconstrained
linear regression as a special case, and it offers bootstrapped estimates of
the standard errors. In the dialog boxes, select Statistics or
Analyze->Regression->Nonlinear, then look in the Options dialog for the
checkbox for bootstrapping. Unfortunately, you have to go through the
trouble of defining the linear equation and giving starting values for the
parameters (power comes at a price).
Another option should you want to consider it would be to create
bootstrapped samples and run them through REGRESSION or GLM and save the
coefficients to work with yourself. The following SPSS commands produce
bootstrapped samples (you have to change nsamples and ncases to the desired
number of samples and the appropriate number of cases).
COMPUTE ID=$CASENUM .
SAVE OUTFILE 'BOOTDATA.SAV'.
INPUT PROGRAM .
LOOP SAMP=1 to nsamples.
LOOP V = 1 to ncases.
COMPUTE ID=TRUNC(UNIFORM(ncases)) + 1.
END CASE.
LEAVE SAMP.
END LOOP.
END LOOP.
END FILE.
END INPUT PROGRAM .
SORT CASES BY ID .
MATCH FILES / FILE * / TABLE 'BOOTDATA.SAV' / BY ID .
SORT CASES BY SAMP.
SPLIT FILE BY SAMP.
David Nichols
Principal Support Statistician and
Manager of Statistical Support