Blavaan: Credible Intervals for Regression Paths (Bayes Path Models)

Question

Is it possible to get credible intervals for regression paths using Blavaan package? Or, p-values for the regression estimates?

R Code:

###------------Run Model: 1A----------------###
Model1A <- '
 C_ERC_ER ~ Valid + S_CCNES 
 S_CCNES ~ C_attachment
 Valid ~ C_attachment
 C_ERC_ER ~ C_attachment
 #Estimating variances of exogenous variables
 C_ERC_ER ~~ C_ERC_ER
 S_CCNES ~~ S_CCNES 
 Valid ~~ Valid
Estimating covariances of exogenous variables
S_CCNES ~~ Valid
'
#Estimating residual variances for endogenous variables
###---evaluate model fit---###
fit1A <- bsem(
  Model1A, data=data, n.chains = 2, 
  burnin = 1000, sample = 1000, 
  target = "stan")
coef(fit1A)
Output:
> summary(fit1A)
lavaan 0.6-11 ended normally after 1000 iterations
Estimator                                      BAYES
  Optimization method                           NLMINB
  Number of model parameters                        12
                                              Used       Total

Number of observations                           117         118
  Number of missing patterns                         2
Model Test User Model:
Test statistic                              -477.860
  Degrees of freedom                                NA
Test statistic                                 0.499
  Degrees of freedom                                NA
Parameter Estimates:
Regressions:
                   Estimate
  C_ERC_ER ~

    Valid            -0.551
    S_CCNES           0.689
  S_CCNES ~

    C_attachment      0.002
  Valid ~

    C_attachment     -0.000
  C_ERC_ER ~

    C_attachment      0.048
Covariances:
                   Estimate
 .S_CCNES ~~

   .Valid             0.007
Intercepts:
                   Estimate
   .C_ERC_ER          5.814
   .S_CCNES           5.404
   .Valid             0.742
Variances:
                   Estimate
   .C_ERC_ER          6.144
   .S_CCNES           0.706
   .Valid             0.076

Answer 1

As @Terrence says, this could be related to convergence issues. Since you are using stan I would suggest first looking at the $\hat{R}$ statistic. $\hat{R}$ > 1.1 are problematic, and for more information see this.

Additionally, as @Terrence says credible intervals should be printed automatically when you apply summary() to a blavaan object. Though when I have used the blavaan R package in the past, I use the highest posterior density region (HPD) - one of many ways to construct Bayesian credible intervals (see this post for more information). In the code chunk below I calculated 95% percentile interval (which, as noted by @Terrence is not always equivalent to the HPD) using a dataset that can be found in the blavaan package (also see this post for a discussion of the 95% HPD in the context the blavaan package).

## Packages 
library(lavaan)
library(blavaan)
## The industrialization and Political Democracy Example
## Bollen (1989), page 332
model <- '
  # latent variable definitions
     ind60 =~ x1 + x2 + x3
     dem60 =~ y1 + a*y2 + b*y3 + c*y4
     dem65 =~ y5 + a*y6 + b*y7 + c*y8
regressions
dem60 ~ ind60
dem65 ~ ind60 + dem60


residual correlations
y1 ~~ y5
y2 ~~ y4 + y6
y3 ~~ y7
y4 ~~ y8
y6 ~~ y8

'
unique priors for mv intercepts; parallel chains
fit <- bsem(model, data=PoliticalDemocracy,
            dp=dpriors(nu="normal(5,10)"))
fitIndices <- blavFitIndices(fit)
Calculating HPD 95% CI
stdpost <- standardizedPosterior(fit)
apply(stdpost, 2, quantile, c(.025,.975))