We develop a method that allows inference on parameters in log-linear models of the relative risk of disease given an individual's haplotypes, that can be used to analyze case-parent trio data. Our methods are robust to population stratification and can also be used for inference on the effect of interactions between haplotypes and environmental covariates. We compare our results with the family-based association test (FBAT) of Horvath et al. ([2004] Genet. Epidemiol. 26:61-69), and discuss when marginal tests, such as those available in FBAT, can be misleading. Our approach generalizes previous results of Allen et al. ([2005] Biometrika 92:559-571), allowing for missing genotype data and haplotype x environment interactions. Additional computational simplifications are also discussed.