Application of time series analysis to improve the validity of Immersive virtual environments for collecting occupant thermal state and adaptive behavioral intention data

Immersive virtual environments (IVEs) are increasingly used to gather occupant perceptions and behavior data for various building designs. However, the use of IVEs for collecting occupant thermal state and thermal adaptive behavior data is still emerging and requires validation. A few studies have attempted to validate thermal-based IVE experimental protocols by statistically comparing central tendency and dispersion measures such as means and standard deviations of thermal state and adaptive behavior responses between IVE and in-situ environments. However, validation methods solely relying on means and standard deviations have limitations. For example, they fail to account for the fluctuations in experimental conditions over time, which potentially contribute to the observed differences in individual paired responses between IVE and in-situ environments. To address these limitations, the authors of this study propose a validation method that identifies the time series factors responsible for the differences in individual paired thermal states and thermal adaptive behavioral intention responses. Ninety-six participants took part in a study that collected their thermal state and thermal adaptive behavior intention data in a climate chamber's virtual and real environments. In addition, data of influential factors like indoor air temperature and skin temperature were also collected. Time series analysis was applied to extract several time series factors from the indoor air and skin temperature data. These factors and the thermal state and behavior intention responses were then modeled as the absolute differences (delta/Δ) between the individual participant pairs in the virtual and the in-situ environments. Regression models were developed to find associations between the delta of the time series factors and the delta of the thermal states and adaptive behavior intentions. The results indicate that certain time series factors, including differences in the timing of peak and trough occurrence in relation to the end of the vote response, differences in the total time of bursts, and variations in irregular patterns of indoor air and skin temperature within virtual and in-situ environments are associated with the participants' thermal sensation, acceptability, comfort, and behavioral intention differences between virtual and real environments. Thus, regulating the temporal fluctuations in the influential factors within an IVE experimental protocol may potentially minimize the discrepancies in participants' thermal state responses between IVE and in-situ environments, thereby improving the ecological validity of IVE experiments on an individual level.