The objective of this study is to analyse uncertainties in building simulation through a probabilistic approach. The EnergyPlus computer programme was used to assess a model of a low-income house located in Florianopolis, southern Brazil. Uncertainties of the user behaviour and physical parameters were obtained through literature review and field survey, respectively. The Latin hypercube was used to create a sample of uncertainties to analyse the thermal performance and energy consumption of the house. A sensitivity analysis using the Standardised Regression Coefficients was performed to obtain the most important parameters in the thermal performance and energy consumption. The results showed that the relative deviation with 95% confidence ranges from 6.6% to 21.5% on the degree-hour for heating and cooling, and from 19.5% to 43.5% on the energy consumption for heating and cooling, respectively. These percentage values are the uncertainties found in each analysis. The specific heat and thickness of materials, solar absorptances, ground temperatures, and the albedo (physical uncertainties) and the occupancy schedules, the equipment power and the number of occupants (user behaviour uncertainties) were important parameters in the analysis. Thus, such parameters must be measured or estimated by more precise methods as they most influence the simulation results. (C) 2014 Elsevier B.V. All rights reserved.