Charlotte Recapet

Dear all, I am trying to model a covariance matrix prior as a mixture of Wishart's distributions, following Mulder & Perrichi 2018 (10.1214/17-BA1092 ). The prior specification however produces an error (whereas the specification for random effects does not): Error in node Psi_res Unable to find appropriate sampler Here are extracts of the model, where priorScale is a 8x8 scale matrix provided with the data. # Likelihood y[i,1:8] ~ dmnorm.vcov(mu[i,1:8], Cov_res) [...] for (i in 1:8){ for (j in 1:8){...

Dear all, I am trying to model a covariance matrix prior as a mixture of Wishart's distributions, following Mulder & Perrichi 2018 (10.1214/17-BA1092 ). The prior specification however produces an error (whereas the specification for random effects does not): Error in node Psi_res Unable to find appropriate sampler Here are extracts of the model, where priorScale is a 8x8 scale matrix provided with the data. # Likelihood y[i,1:8] ~ dmnorm.vcov(mu[i,1:8], Cov_res) [...] for (i in 1:8){ for (j in 1:8){...

Hello Ayca, It looks like the equivalent of R function combn does not exist in JAGS, so the valvariable cannot be created from i and j. You should create the combination manually within your JAGS model. Hope that helps!

I indeed tried fixing initial values (similar to your simpler solution), but these did not fulfill the function constraints with my full dataset. The first solution you propose, directly making initial values consistent with observations, however solved the problem! Thanks for your help!

Dear all, I am exploring the possibility to use the dinterval function to code threshold models. When using fixed thresholds, the model runs alright, but when using a random threshold, even with strict constraints to truncate it within the range of fixed thresholds, I get a "Node inconsistent with parents" error. I attached toy data to reproduce the error using the code below. Any help to understand what might be the problem here would be helpful, thanks! Charlotte sink("src/Toy model.md") cat("...

Dear all, I am exploring the possibility to use the dinterval function to code threshold models. When using fixed thresholds, the model runs alright, but when using a random threshold, even with strict constraints to truncate it within the range of fixed thresholds, I get a "Node inconsistent with parents" error. I attached toy data to reproduce the error using the code below. Any help to understand what might be the problem here would be helpful, thanks! Charlotte sink("src/Toy model.md") cat("...