The Effect of Energy Efficiency Measures on the Prevalence, Frequency and Intensity of Overheating in the EFUS2017 Dwellings: Statistical Analysis

This supplementary information contains worksheets of the statistical analyses that underpin the narrative and charts presented in the following paper: Lomas KJ, Li M and Drury P (2024) How do energy efficiency measures affect the risk of summertime overheating and cold discomfort? Evidence from English homes, Energy Policy, Volume 188, 15pp, 114108, ISSN 0301-4215, https://doi.org/10.1016/j.enpol.2024.114108.<b>Analysis method</b>The samples of 616 living rooms and 591 main bedrooms were first divided into those in houses (506 living rooms, 496 bedrooms) and flats (110 living rooms, 95 bedrooms). Each of these 4 samples was then divided according to 5 energy efficiency variables (loft insulation thickness, predominant wall type, whether walls were insulated, whether windows were all double-glazed, EPC energy efficiency rating band) and floor level for those in flats (whether on the top floor of a building, and whether the entrance was above ground floor level).The prevalence, frequency and intensity of overheating experienced by each room during the summer (May to September inclusive) of 2018 was calculated (see the associated paper for definitions of these metrics and the threshold temperatures applied). Frequency and intensity of exceedance of lower temperatures associated with cold discomfort and thermally neutral conditions was also assessed.Statistical testing was then undertaken to assess the significance of between-categories differences in prevalence of overheating and frequency and intensity of threshold exceedance, e.g., comparing living rooms in houses with and without insulated walls. Differences in overheating prevalence were assessed using chi-square tests, while those for frequency and intensity of threshold exceedance were assessed using independent t-tests. For frequency and intensity of exceedance, Box-Cox transforms were required to produce approximately normally distributed data; as a result, only rooms having non-zero hours of exceedance were included in the analyses of frequency and intensity.