Graphics Reference
In-Depth Information
Figure
.
.
A graphical summary (produced automatically by the R package R
WinBUGS Sturtz et al.
(
)) of the estimated scalar parameters in the social networks model. Another alternative is to plot
density estimates or histograms of individual scalar parameters; these can be useful, but are di
cult to
apply to vector parameters. In contrast, the right side of our display here permits an immediate
understanding of the inferences for the vectors α, β, ω (another option would be to use parallel
coordinate plots for each vector parameter). his graph is an ine
cient way of displaying inferences;
given that convergence has approximately been reached, only the right side of the display is necessary.
However, we include it as an example of a quick inferential summary which, for all its simplicity, is still
far more informative than a table of parameter estimates and standard errors
display the posterior mean estimates of a
i
as two separate histograms by dividing α
i
into men's and women's estimates. See Fig.
.
.
However, a histogram derived from the posterior means of hundreds of parame-
ters is still only a point estimate; we also want to visualize the posterior uncertainty
