| plotModelsize {BMS} | R Documentation |
Plots posterior and prior model size distribution
plotModelsize(bmao, exact = FALSE, ksubset = NULL, include.legend = TRUE, do.grid = TRUE, ...)
bmao |
a 'bma' object (cf. bms) |
exact |
if TRUE, then the posterior model distribution is based on the best models of bmao and their marginal likelihoods;if FALSE (default) then the distribution is based on all encountered models and their MCMC frequencies (cf. 'Details' in coef.bma) |
ksubset |
integer vector detailing for which model sizes the plot should be done |
include.legend |
if TRUE, a small legend is included via the low-level command legend |
do.grid |
if TRUE, a grid is added to the plot (with a simple grid()). |
... |
parameters passed on to matplot with sensible defaults |
As a default, plotModelsize plots the posterior model size distribution as a blue line, and the prior model distribution as a dashed red line.
In addition, it returns a list with the following elements:
mean |
The posterior expected value of model size |
var |
The variance of the posterior model size distribution |
dens |
A vector detailing the posterior model size distribution from model size 0 (the first element) to K (the last element) |
Martin Feldkircher and Stefan Zeugner
See also bms, image.bma, density.bma, plotConv
Check http://bms.zeugner.eu for additional help.
data(datafls) mm=bms(datafls,burn=1500, iter=5000, nmodel=200,mprior="fixed",mprior.size=6) #plot Nb.1 based on aggregate results postdist= plotModelsize(mm) #plot based only on 30 best models plotModelsize(mm[1:30],exact=TRUE,include.legend=FALSE) #plot based on all best models, but showing distribution only for model sizes 1 to 20 plotModelsize(mm,exact=TRUE,ksubset=1:20) # create a plot similar to plot Nb. 1 plot(postdist$dens,type="l") lines(mm$mprior.info$mp.Kdist)