Hello Everyone:

My question relates to estimating a bivariate ordered model with random coefficients in both equations using Roodman's (2011) cmp. I wish to have the random coefficients uncorrelated within equation and across the two equations, but I do want to have the correlation between the residuals from the two equations. This is what I tried (based on my previous experience with cmp and it worked back then):

Code:
capture drop id
gen id = _n
cmp setup
cmp(DV1 =A B C D E F || id: C D, noconstant cov(indep)) ///
(DV2 = G H I J K A C || id: G H, noconstant cov(indep)), ///
ind($cmp_oprobit $cmp_oprobit) redraw(10, scramble) nonrtolerance difficult cov(indep .)
When I run this, the software attempts to estimate the intra-equation correlations (despite me requesting zero intra-equation correlations) and gives the following error:
"equation atanhrho_1_2_1_1 not found
invalid syntax"

Question: How should I modify the above command to have uncorrelated random coefficients within and across equations BUT to have cross-equation correlation between the residuals? Based on my previous experience with cmp, the above specification of cov at the end and in the specification of the hierarchical portions was fully working in the past for bivariate probit model. Now, strangely even for that same bivariate probit model it is not working and I am unable to reproduce what I produced back then. Perhaps something has changed in the updated versions of cmp? Many thanks for your valuable attention! -behram