monitor {R2WinBUGS} | R Documentation |
Special summary statistics of the WinBUGS output - intended for internal use only
monitor(a, n.chains, trans = NULL, keep.all = FALSE, Rupper.keep = FALSE) conv.par(x, n.chains, Rupper.keep = TRUE)
x |
|
a |
a n * m * k array: m sequences of length n , k variables measured |
n.chains |
number of Markov chains |
trans |
a vector of length k : "" if no transformation, or "log" or "logit"
(If trans is NULL , it will be set to "log" for parameters that
are all-positive and 0 otherwise.) |
keep.all |
if FALSE (default), first half of a will be discarded |
Rupper.keep |
if FALSE , don't return Rupper |
See the well documented code for details.
for monitor
:
output |
list of "mean","sd", quantiles ("2.5%","25%","50%","75%","97.5%"), "Rhat" if n.chains > 1 ,
"Rupper" if (Rupper.keep == TRUE) && (n.chains > 1) , and "n.eff" if n.chains > 1 |
quantiles |
emipirical quantiles of simulated sequences |
confshrink |
estimated potential scale reduction (that would be achieved by continuing simulations forever) has two components: an estimate and an approx. 97.5% upper bound |
n.eff |
effective sample size: m*n*min(sigma.hat^2/B,1) .
This is a crude measure of sample size because it relies on the between
variance, B , which can only be estimated with m degrees of freedom. |
The main function to be called by the user is bugs
.