R-hat statistic

[Sanne Appels]

Hello all,

I'm currently using Blavaan and was wondering about the default Rhat statistic that Blavaan uses. Does this entail the traditional R-hat, the split R-hat from Gelman et al. 2013 or the rank normalized split-R-hat statistic as proposed by Vehtari et al. 2021?


Gelman, A., Carlin, J. B., Stern, H. S., Dunson, D. B., Vehtari, A., and Rubin, D. B. (2013). Bayesian Data Analysis, third edition. CRC Press. MR3235677. 668, 669, 670, 672, 673, 675, 676, 687)

Aki Vehtari. Andrew Gelman. Daniel Simpson. Bob Carpenter. Paul-Christian Bürkner. "Rank-Normalization, Folding, and Localization: An Improved ˆR for Assessing Convergence of MCMC (with Discussion)." Bayesian Anal. 16 (2) 667 - 718, June 2021. https://doi.org/10.1214/20-BA1221

Many thanks for the response!

Kind regards,

Sanne Appels

[Ed Merkle]

Sanne,

Good question. blavaan displays the Rhat values that the estimation software (Stan or JAGS) reports. Assuming you are doing estimation in Stan (the default for blavaan), then this post might be helpful:

https://discourse.mc-stan.org/t/how-rstan-compute-rhat/38322/3

Based on that post, it looks like the Rhat from blavaan's summary() and blavInspect() commands is "v3". If you wanted the newer "v5", it could be obtained via

rstan::monitor(blavInspect(fit, 'mcobj'))

And to make sense of the parameter names in the output, do

pt <- parTable(fit)

with(pt[pt$free > 0,], cbind(paste0(lhs, op, rhs), pxnames))

Ed

[Sanne Appels]

Hello Ed,

Thank you for the fast response, the post, and further clarification. I think v3 will work for me, but I will also look into the v5 version with the code you provided.

Kind regards,

Sanne