Causal thinking: Uncovering hidden assumptions and interpretations of statistical analysis in building science

Causal thinking emphasizes the understanding of asymmetric causal relationships between variables, requiring us to specify which variable is the cause (independent variable) and which is the effect (dependent variable). Reversing the causal relationship direction can lead to profoundly different assumptions and interpretations. We demonstrate this by comparing two linear regression methods used in thermal comfort data analysis: Approach (a), which regresses thermal sensation votes (y-axis) on indoor temperature (x-axis), and Approach (b), which does the reverse, regressing indoor temperature (y-axis) on thermal sensation votes (x-axis). From a correlational perspective, two approaches may appear interchangeable, but causal thinking reveals substantial and practically significant differences between them. Approach (a) aligns with most laboratory studies and considers occupants' thermal sensations as responses to indoor temperature. In contrast, Approach (b), rooted in adaptive comfort theory, suggests that thermal sensations can trigger behavioral changes, which in turn alter indoor temperature. Using the same data, we found that Approach (b) leads to what we call a 'preferred zone', which is 10 °C narrower than the conventionally derived comfort zone using Approach (a). The 'preferred zone' might be interpreted as the thermal conditions that occupants are likely to choose when they have significant control over their personal and environmental thermal settings. This finding has important implications for occupant comfort and building energy efficiency. We highlight the importance of integrating causal thinking into correlation-based statistical methods, which have been prevalent in building science research, especially given the increasing volume of data in the built environment.