An alternative statistical approach to estimate the level of airtightness of existing residential buildings: Influencing factors from measured data

This is the authors' abstract. We don't add key points for this paper.

• A predictive model to estimate the envelope airtightness of existing dwellings. • Cross-validated GLIM model with enhanced predictive capacity. • GLIM model involving 13 main effects and 4 interactions. • The model allows to transcend region-specific applicability. Estimating the level of airtightness of a building can offer valuable information for energy performance simulation tools or decision-making during retrofitting processes. However, it remains a challenge given the great variability of the variables involved, the complexity of addressing some of these variables, and some context-specific features. Based on previous research in this direction, this paper proposes an alternative predictive model based on Generalized Linear Models (GLIM) and validated using cross-validation that involves 13 main effects and 4 interactions. This leads to a substantial enhancement in predictive capacity, accounting for nearly 50% of the response variability. A detailed set of variables fully described offers the opportunity to transcend region-specific applicability and opens a window for other populations. The model provides more reliable estimates of airtightness and expands its applicability to a broader range of construction conditions, while maintaining the statistical significance of its predictors and achieving a satisfactory fit.