Open qscx1364 opened 1 year ago
I see that the authors said glmqap was soft deprecated, but the dyadic_brm/ or the multi-membership brms would really cost a tremendous time for model fitting and always cannot converge, (I have used the parallel procedure to have nearly 100% CPU use of my PC ) ...... OMG...... (I think that a network of 400x400 with many zeros is not that big?) ......however.... Maybe I would prefer to use the glmqap though it was a solution not that good.
I have some questions in using the dyadic bayesian regression models to replace the deprecated QAP.
(1) I have got warning of "Response matrix is not symmetric; this function is designed for undirected networks." I want to know that could I still use the results based of dyadic_brm(), even my response matrix is directed (a 400×400 asymmetric matrix of population migration). (2) or ... I have to use the GLMQAP to deal with directed network? since the authors have pointed out problems of QAP and reccommended to use dyadic brm (3) Should the matrices of independent variables also be symmetric?