docs.javahelp.manual.boxes.search.images.html Maven / Gradle / Ivy
Search Algorithms: iMAGES
Search Algorithms: iMAGES
iMAGES (Model Averaging GES) is the GES algorithm, using a model
averaging BIC score that lets the algorithm search over multiple data
sets simultaneously. For a description of the GES algoritihm, see GES.
iMAGES takes as input a list of continuous data sets D1 through Dn,
all of which contain the same varibles, in the same order, and all of
which contain the same number of cases. (The discrete case has not
been tested.
The BIC score of iMAGES, for a given structural hypothesis in the
form of a DAG G, is the average of the BIC scores for G with respect
to each data sets Di, i = 1,...,10. The search proceeds in the order
of GES--that is, calculating the score of the empty graph, then
finding the edge that most increases the average BIC score and adding
that, and so on.
A "discount penalty" may be applied, which multiplies the penalty
part of the BIC score by the given number (or, equivalently, divides
the likelihood part by the given number--a simpler interpretation when
nonlinear functions are involved in generating the data).
Figure 1 shows the current dialog for this search, which includes
an untested discrete sections.
Figure 1