Derek posted on Monday, January 28, 2019 - 5:52 pm
Thanks in advance. I have a couple of general questions about cross-lagged models (using Mplus).
1) I am interested in the reciprocal relationship between state alcohol policy score (a continuous variable) and the number of mass violence incidents in that state (a count variable) between, for example, 2009 and 2018. Can I do a cross-lagged model between a continuous variable and a count variable? Any caveats I should be aware of?
2) If the answer is "yes" to the above question, can I add additional covariates in the cross-lagged model? For example, I want to know the reciprocal relationship between state alcohol policy and the number of mass violence incidents while controlling for state-level crime rates.
If possible, please direct me to the references (e.g. pages in the Mplus manual or other papers).
1) Having a count and continuous DV at the same time is not a problem. But when the count variable is then an IV for the next time point, you have a problem. There is no natural underlying continuous latent response variable Y* for counts like there is for binary and ordinal variables. So you would end up using the same count variable modeled as a count in its DV equation and as a continuous variable in its IV equation - not a good situation.
So perhaps a better way is to create an ordinal variable out of the count variable - and perhaps this is natural also from a substantive point of view, using categories like (just as an example): category 0 for count = 1-5, category 1 for count 6-15, category 2 for count > 15. Then you have a Y* variable to work with as both a DV and an IV. And you can use WLSMV or Bayes to estimate.