Summary function for accessing the Stanfit object

stan_beginer · January 22, 2021, 5:02am

Hi,

I am trying to access the content of Stanfit object using summary function given in this post: Accessing the contents of a stanfit object and I am wondering whether I could find the sourse code for the summary function. Currently I found that after applying the summary function, some of the parameters have values in ‘mean’ but NaN in their ‘se_mean’, ‘n_eff’ and ‘Rhat’ and I am wondering whether this might be a problem due to the summary function.

Thx!

ahartikainen · January 22, 2021, 7:30am

Hi, source code for RStan summary is here

github.com

stan-dev/rstan/blob/ae36e6f9842e9e16ef79700ce86a122bfa9c5b10/rstan/rstan/R/stanfit-class.R#L495


            sssf <- unlist(sss2, use.names = FALSE) 
  
            n2 <- object@sim$n_save[1]  ## assuming all the chains have equal iter 
            if (!inc_warmup) n2 <- n2 - object@sim$warmup2[1] 
            dim(sssf) <- c(n2, object@sim$chains, length(tidx)) 
            cids <- sapply(object@stan_args, function(x) x$chain_id)
            dimnames(sssf) <- list(iterations = NULL, chains = paste0("chain:", cids), parameters = tidxnames)
            sssf 
          })  
setMethod("summary", signature = "stanfit", 
          function(object, pars, 
                   probs = c(0.025, 0.25, 0.50, 0.75, 0.975), use_cache = TRUE, ...) { 
            # Summarize the samples (that is, compute the mean, SD, quantiles, for 
            # the samples in all chains and chains individually after removing
            # warmup samples, and n_eff and split R hat for all the kept samples.)
            # 
            # Returns: 
            #   A list with elements:
            #   summary: the summary for all the kept samples 
            #   c_summary: the summary for individual chains.

And for monitor here

github.com

stan-dev/rstan/blob/ae36e6f9842e9e16ef79700ce86a122bfa9c5b10/rstan/rstan/man/monitor.Rd

\name{monitor}
\alias{monitor}
\alias{print.simsummary}
\alias{[.simsummary}
\title{Compute summaries of MCMC draws and monitor convergence} 
\description{
  Similar to the \code{print} method for \code{stanfit} objects, but \code{monitor}
  takes an array of simulations as its argument rather than a \code{stanfit}
  object. For a 3-D array (iterations * chains * parameters) of MCMC draws, 
  \code{monitor} computes means, standard deviations, quantiles, Monte Carlo 
  standard errors, split Rhats, and effective sample sizes. By default, half of 
  the iterations are considered warmup and are excluded.
}

\usage{
monitor(sims, warmup = floor(dim(sims)[1]/2), 
        probs = c(0.025, 0.25, 0.5, 0.75, 0.975), 
        digits_summary = 1, print = TRUE, ...)
\method{print}{simsummary}(x, digits = 3, se = FALSE, ...)
\method{[}{simsummary}(x, i, j, drop = if (missing(i)) TRUE else length(j) == 1)

This file has been truncated. show original

ahartikainen · January 22, 2021, 7:33am

Can you check your parameters, are all the numbers the same for some parameters?

How many chains do you have?

stan_beginer · January 22, 2021, 4:13pm

Hi,

I have four chains with default iterations in Rstan. I have applied the ‘extract’ function to check the draws of parameters, and for some parameters having this issue they all have the same values (all 0s), while for other parameters having this issue they have different numbers (many 0s plus some small numbers like 3e^-300, 7e^-150).

Thx!

ahartikainen · January 22, 2021, 4:23pm

Ok, then your model probably has some problems (model definion; e.g. missing priors) or your model is not suitable for your data.

Can you share your model?

stan_beginer · January 22, 2021, 4:36pm

Sure and my model is included in one of my other posts here:
Undefined values in generated quantities block. The problem happened for generated quantities like ‘numerator’ and ‘denominator’.

Thx!

Topic		Replies	Views
NaN values in Stanfit summary object General fitting-issues	0	273	May 4, 2023
Assignment version of extract in Rstan RStan rstan	1	816	February 5, 2019
Rstan output structure -- coerce samples back to 'rich' structure (arrays, matrices, etc) General	2	603	January 15, 2018
Rstan::read_stan_csv chokes on NaN RStan	3	826	July 25, 2018
What are the rstanarm::stan_glmer Error Term standard deviations reporting? Modeling rstanarm	1	481	March 16, 2021

Summary function for accessing the Stanfit object

Related topics