An approach for a statistical model for the user behaviour regarding window ventilation in residential buildings

In this paper a user model for the use of windows based on measurements in residential buildings is developed. Detailed literature review provides an overview of existing user models. Most of them base on measurements in office buildings and their validity for residential buildings is questionable. All models are based on different kinds of statistical methods.  After detailed assessment of these methods and the analysis of their applicability to the measurements, a new methodological approach is applied, to take the auto-correlation in the longitudinal data into account. The method of Generalized Estimating Equations (Liang and Zeger 1986) for modelling user behaviour regarding natural ventilation is chosen. By using this class of models, the auto-correlation of the non-normally distributed data can be implied in an adequate way. A binary and a continuous response variable regarding the user behaviour is modelled involving various explanatory variables with different scale levels. A model for predicting the probability of window opening and a model for predicting the window open duration is developed. Both statistical models show positive goodness of fit results by a cross-validation using lift plots.