Using causal inference to avoid fallouts in data-driven parametric analysis: a case study in the architecture, engineering, and construction industry