I'm running an analysis looking at macro-predictors of some different processes. Because the macro predictors (higher levels stuff like average years of education, labor force participation, GDP, etc.) are basically a causal salad/soup, I started by running a multiverse analysis. I ran a regression for all possible combinations of my predictors, and then plot the t-values. I then pull the predictors whose t-values never cross zero (so they are always in the same direction) and also cross the significance threshold at some point in their distribution.

The predictors I pulled I then am putting into SEM model. BUT - one of my predictors goes from being ALWAYS negative in the multiverse analysis to positive in the SEM. This doesn't make sense to me. Is this something that might be expected? If so, how do I make sense of this? Or is this as weird as I think it is? If this is as weird as I think this is - what are some things I should try troubleshooting in this analysis?



Your Answer

By clicking “Post Your Answer”, you agree to our terms of service and acknowledge you have read our privacy policy.

Browse other questions tagged or ask your own question.