The gaper discusses problems connected with the inclusion of stochastic factors in deterministic models of indoor air quality (IAQ). Three different methods are shortly presented: quasi-dynamic multizone modelling with generation of input data time series; multizone modelling based on the theory of stochastic differential equations; and Monte Carlo simulation with independent random generation of stochastic parameters. The described methods are compared using a computer simulation of carbon dioxide concentration in a simple two-compartment office. The comparison of simulation results shows that the way in which stochastic disturbances are included in the models does not have an important influence on mean value of predicted carbon dioxide concentration. At the same time, the analysis of standard deviations indicates that the method of disturbance generation and its later incorporation into the IAQ models have a great influence on the probability distribution of estimated concentrations. Finally, there is a discussion on the main advantages and disadvantages of each of the proposed methods.