In your article, you stressed the importance of the choice of the initial values like:
Dbeta =ginv(X)%%Y # ML solution
Dhat=apply(Dbeta,MARGIN=1,FUN=sd)# extract SD
beta=Dbeta/Dhat
// Decompose Beta into L and Ahat
tmp=qr(t(beta))
iL=t(qr.R(tmp))
iL=iL[,1:data_list$k]
iQ=t(qr.Q(tmp))
iQ=iQ[1:data_list$k,]
iL = iL/sqrt(data_list$kdata_list$c)
iAhat = iQsqrt(data_list$kdata_list$c)
// flip iB so diags are positive
sgn=t(sign(iL[diag_idx]))
iL = iL%%diag(c(sgn))
iAhat = diag(c(sgn))%%iAhat
How can I enter the above list to stan? I tried to do as below but failed:
fit ← stan(file = “reduced_rank_002.stan”, init=list(L=iL,Ahat=iAhat), data = data_list, warmup = 5000, thin=1, seed = 2, iter=10000)
I am not familiar with stan. Thank you for your help.