Hello,

I am running a series of multinomial logistic regressions that use the same outcome variable and control variables, but different focal predictors, e.g.:

y= x1 + controls
y= x2 + controls
y= x3 + controls
y= x4 + controls
y= x5 + controls

I would like to run a test to address potential correlation of error terms in these models. I have looked into SUR techniques (specifically "suest" after storing the estimates in Stata), but these seem to be utilized when regressing different outcomes on the same predictors, e.g.:

y1 = x + controls
y2 = x + controls
y3 = x + controls
y4 = x + controls
y5 = x + controls

Is there a SUR-like technique that can be used to address the correlation of error terms for multinomial logistic regressions using the same outcome but different predictors?

Thank you,
Sarah