Mario posted on Monday, January 11, 2016 - 11:26 pm
I am wondering if it is possible to run a SEM model where for some dependent variables I have a complete data set (e.g. the number of offspring per female) and for the others only a subset of this data set (e.g., male offspring body mass, which means that the data is missing for all female offspring). Can the model estimate the different associations while ignoring the missing data? or should I remove first all the rows that are not complete?
This is handled by the default "FIML", that is, ML under MAR. Don't remove rows.
Mario posted on Sunday, January 24, 2016 - 5:09 am
Thanks for the answer. We are wondering whether we can use MLR in the same way since our dependent variables are not normally distributed. Is also "FIML" the default? should I remove rows in this case?