The study employed Tobit models corrected for heteroscedasticity and applied to Canadian household survey data to analyze household consumption of food away from home (FAFH) by type of establishment and type of meal. Results suggest that failing to correct for heteroscedasticity causes most model parameters to be overestimated. Other results indicated that the influence of socioeconomic and demographic variables on FAFH expenditures were conditional on both the type of meal and type of establishment.