You say you have the actual response values and all the inputs, which suggests you can simply refit the model in PROC LOGISTIC and use the OUTPUT statement to generate the regression diagnostics of interest.  ... View more

Sorry... I was thinking of the equivalence between log-linear models and logistic models when I mentioned an "identifiable response". Ignore my comments about a response in my reply. ... View more

See the new SAS/STAT procedure, PSMATCH.  ... View more

These are really the same thing when there is an identifiable response variable. The log-linear model from earlier days was used to model cross-classified data and the relationships among the various categorical variables. There might not be a single "response" variable. This was originally available in SAS via the LOGLIN statement in PROC CATMOD. But the Poisson model does the equivalent and, these days, this generalized linear modeling approach is the preferred analysis. For instance, the "Log-Linear Model, Three Dependent Variables" in the CATMOD documentation can be reproduced in GENMOD using a Poisson model by using the cell counts as the response and specifying the three variables as CLASS predictors. ... View more

As noted in the NLEstimate macro description of the DF= option: "The degrees of freedom for testing a linear combination of parameters in a linear model would typically be the number of observations used in fitting the model minus the number of parameters estimated in the model – essentially, the error degrees of freedom."  So for your example with 120 observations and a single binary predictor resulting in two model parameters being estimated, you could use 118 degrees of freedom. But note that large sample Wald statistics are often used, as they are in procedures like GENMOD, which are chi-square with 1 df. So, you could get the Wald statistic as (estimate/stderr)**2 and its p-value is 1-probchi((estimate/stderr)**2, 1) .  At df as large as 118, you probably won't see much difference in the two p-values. ... View more