I have data for several outcomes which I would like to analyze using a multivariate mixed effects model. My understanding is that if I "melt" the data to long format, I can include a categorical indicator for each response variable. Some links to this approach are:
https://rpubs.com/bbolker/3336
How to interpret coefficients of a multivariate mixed model in lme4 without overall intercept?
From my reading, I understand it necessary to include random slopes for the response variable indicators to correctly account for correlations. However, I wondered if a case could be made for NOT including these random slopes; for example, if the model fails to converge, the model fit (e.g. likelihood ratio test, graphical inspection of residuals etc.) is superior in the model WITHOUT the random slopes. Any advice is warmly appreciated.
