NAs for the rhat values of Cholesky factor

Modeling
,
1

Hi all,

I’m new to brms and am trying to fit my first ever models. I was wondering if it is normal to have NAs for the rhat values of the Cholesky factor? Or does it mean my model has not converged successfully?

See the model specification below and the output from the rhat function. As you can see, the values for L_1[1,1] and L_1[1,2] are NAs. Thank you in advance!


# set prior // generic weakly informative prior 
prior <- c(set_prior("normal(0,1)", class = "b"), # fixed effect 
           set_prior("normal(0,1)", class = "sd")) # random effect 

# run model with replicate as random effect 
mod_daysflow_hmo1 <- brm(days_to_flowering ~ Treatment + (Treatment|Accession) + (1|Replicate), 
                        data = hmo_greenh, 
                        save_pars = save_pars(all = TRUE), 
                        warmup = 2000, 
                        iter = 4000,
                        chains = 4, 
                        thin = 1, 
                        control = list(adapt_delta = 0.999, 
                                       max_treedepth = 10), 
                        seed = 1, 
                        prior = prior)


> rhat(mod_daysflow_hmo1)
                        b_Intercept                         b_Treatment             sd_Accession__Intercept 
                          0.9999061                           1.0002709                           0.9999284 
            sd_Accession__Treatment             sd_Replicate__Intercept cor_Accession__Intercept__Treatment 
                          1.0018551                           1.0010493                           1.0007562 
                              sigma                           Intercept        r_Accession[hmo_1,Intercept] 
                          1.0006724                           1.0018050                           1.0001828 
       r_Accession[hmo_2,Intercept]        r_Accession[hmo_3,Intercept]        r_Accession[hmo_4,Intercept] 
                          1.0014713                           1.0006722                           1.0004134 
       r_Accession[hmo_5,Intercept]        r_Accession[hmo_1,Treatment]        r_Accession[hmo_2,Treatment] 
                          1.0006772                           1.0006099                           1.0016887 
       r_Accession[hmo_3,Treatment]        r_Accession[hmo_4,Treatment]        r_Accession[hmo_5,Treatment] 
                          1.0017980                           1.0010584                           1.0008110 
          r_Replicate[CA,Intercept]           r_Replicate[CB,Intercept]           r_Replicate[HA,Intercept] 
                          1.0002018                           1.0001680                           1.0005006 
          r_Replicate[HB,Intercept]           r_Replicate[MA,Intercept]           r_Replicate[MB,Intercept] 
                          1.0002751                           0.9999867                           1.0011161 
          r_Replicate[WA,Intercept]           r_Replicate[WB,Intercept]                              lprior 
                          1.0006978                           1.0003166                           1.0007380 
                               lp__                            z_1[1,1]                            z_1[2,1] 
                          1.0016656                           1.0002565                           1.0003426 
                           z_1[1,2]                            z_1[2,2]                            z_1[1,3] 
                          1.0013754                           1.0006062                           1.0012323 
                           z_1[2,3]                            z_1[1,4]                            z_1[2,4] 
                          1.0004641                           1.0001609                           1.0002042 
                           z_1[1,5]                            z_1[2,5]                            L_1[1,1] 
                          1.0019945                           1.0001353                                  NA 
                           L_1[2,1]                            L_1[1,2]                            L_1[2,2] 
                          1.0007562                                  NA                           1.0006377 
                           z_2[1,1]                            z_2[1,2]                            z_2[1,3] 
                          1.0002668                           1.0002606                           1.0001975 
                           z_2[1,4]                            z_2[1,5]                            z_2[1,6] 
                          1.0008213                           1.0001702                           1.0005400 
                           z_2[1,7]                            z_2[1,8] 
                          1.0002573                           1.0006483
2

Yes, those values with NA are deterministic. L_{1,1} is always 1 and the rest of that row (including L_{1,2}) are all 0.

3 Likes