The way buildings are operated highly influences energy consumption, and occupants play a significant role in it. However, challenges related to the prediction of occupants’ behavior decrease the potential to accurately predict a building’s energy consumption. Such inaccuracy presents one reason for the gap between measured and simulated energy data. In an effort to bridge this gap, this work aims at creating statistical models to predict occupant behavior, in relation to window and air conditioning operation in mixed-mode office buildings in a humid subtropical climate. The statistical behavioral models developed were based on data collected in an 18-month monitoring campaign conducted in São Carlos, Brazil. Two types of statistical methods were applied: generalized linear mixed and Markov chain models. Model results showed high probabilities of status change in both controls upon arrival and departure. Indoor temperature approximately at 24 °C and outdoor temperature around 23 °C show the lowest probability of either control being activated. In addition, the behavioral patterns presented by the models show an alternating use of the controls, indicating a conscious use of the available equipment.