search for: pdsymmm

...roups to be independent. The default class for pdMat, pdSymm, assumes that ALL random intercept and slope parameters are correlated. The reality is that the intercepts and slopes should be correlated within each subgroup, but these random parameters are independent between the subgroups. Use of the pdSymmm class leads to parameter non-identifiability. What I need is something like the 'value' argument for pdSymm, by which one can specify initial covariance parameter estimates for the random effects. If there were a way to specify that initial covariance values of 0 are to be fixed at 0, then...