A random-effects regression was used because the models included important time-invariant variables, including location, the class of hotel and the size hotel. The models were to examine how important the F&B operating was in hotel business performance in the aspect of occupancy, ADR, RevPAR, GOPPAR, and GOP ratio. The results of the Breusch-Pagan/ Cook-Weisberg test (Breusch & Pagan, 1979) for heteroskedasticity indicated the variance was not homoscedastic (p-value < 0.001) and Ramsey RESET test (Ramsey, 1969) also showed that the errors were autocorrelated (p-value