equiv_classes_hier_hmm.m

The BNL toolbox is a set of Matlab functions for defining and estimating the parameters of a Bayesi

function [equiv_class,param_equiv_class,link]=equiv_classes_hier_hmm(onames,B,D)


%equiv_class: vector. (ordered) nodes having the same CPT have same value
%param_equiv_class: vector.  (ordered) nodes having a CPT governed by the
    %same set of parameters have the same value. CPTs may be different (eg when because
    %different values on covariates)
nr_nodes=length(onames);
obs_nodes=strmatch('Y', onames);
hid_signal_nodes=strmatch('Xsignal', onames) ;
hid_day_nodes=strmatch('Xday', onames) ;    
start_day_node=strmatch('Xday_1', onames) ;
start_signal_nodes=[];
for i=1:B:B*D
    v=eval(['strmatch(''Xsignal_',int2str(i),''',onames,''exact'')']);
    start_signal_nodes=[start_signal_nodes v(1)];
end
    
param_equiv_class=zeros(nr_nodes,1);
param_equiv_class(start_day_node)=1;
param_equiv_class(mysetdiff(hid_day_nodes,start_day_node))=2;
param_equiv_class(start_signal_nodes)=3;
param_equiv_class(obs_nodes)=4;
param_equiv_class(mysetdiff(hid_signal_nodes,start_signal_nodes))=5;

equiv_class=param_equiv_class;
f5=find(equiv_class==5);
equiv_class(f5)=5:length(f5)+4;


%link function for all equiv_classes (multinomial cumulative adjacent)
for i=1:max(equiv_class)
link{i}='multinomial';
end