Multiverse analyses with meta-analytic aggregation can be used to evaluate cross-lagged effects: Examples with anxiety, depression, and aggression

A cross-lagged effect of initial X on subsequent Y when adjusting for initial Y may be spurious due to a correlation with residuals in combination with regression to the mean. Here, we propose that cross-lagged effects can be evaluated by multiverse analyses of the effect of initial X on subsequent change in Y when adjusting for (1) initial Y; (2) subsequent Y, and (3) neither initial nor subsequent Y. These three effects can, then, be meta-analytically aggregated and the meta-analytic effect used for conclusions. We applied this method on data simulated to resemble data used in three recent studies. We present null effects of initial internet gaming disorder, optimism, and social media addiction on subsequent change in anxiety, depression, and overt aggression, respectively. Hence, the present results call conclusions in the original studies into question.