In aqua veritas: Triangulation of cross-lagged effects for improved causal inference

Objective: To show which adjusted cross-lagged and concurrent effects to expect from cross-lagged panel models in the case of true causal increasing and decreasing effects.Design: Full factorial prospective in situ experimental study.Setting: The first author’s kitchen.Participants: Twenty plastic cups.Main outcome measures: Total weight (i.e. including their content) of the cups before and after adding or removing water to/from them. Results: We found the amount of water added to cups to have: (1) A positive effect on the subsequent total weight of the cups when adjusting for their initial weight; (2) A negative effect on the initial weight of the cups when adjusting for their subsequent weight; (3) A positive effect on the subsequent weight - initial weight difference of the cups. Removing water from cups filled with water had the opposite effects.Conclusions: The same combination of signs of effects as when adding or removing water to/from cups should be expected in the case of true increasing or decreasing effects between studied constructs in all areas of research, such as psychology. Estimation of all three of these effects, instead of only the first one, allows stronger inference about causality and is recommended in order to triangulate and scrutinize findings from cross-lagged panel models.