back transforming beta values

[tinah]

I am running the robust design with full heterogeneity with 1 group and 8 covariates:scale, burn, feed, 2002, 2003, 2004, 2005 and sex. When I run a model with the covariates feed and sex, I back transform the beta values to distinguish the survival estimate for say male-unfed rats or one of the other combinations using the only the Survival betas generated by Mark. However, when I run the analysis where I pick the "user specified covariate values, and pick the covariates specify male and unfed, the actual survival estimates that are generated are different than those that I back transformed by hand. How is Mark back transforming the beta values. Should I assume that I am making a calculation error somewhere and trust Mark, or trust myself? Usually I do not trust myself. Thanks.