Far East Journal of Theoretical Statistics
Volume 13, Issue 2, Pages 127 - 133
(July 2004)
|
|
A NOTE ON THE RELATION BETWEEN MULTIVARIATE STRICT COLLAPSIBILITY AND CONDITIONAL INDEPENDENCE IN DIRECTED LOGLINEAR MODELS
E. M. D. Aris (France)
|
Abstract: In this note, it is shown that for Directed Loglinear Models, the simple use of conditional independence allows to find the maximum set of variables that can be omitted during the estimation of the parameters as also provided by the use of the multivariate strict collapsibility property. As a consequence, for Directed Loglinear Models, reductions obtained by means of strong or strict collapsibility properties coincide. |
Keywords and phrases: logit model, collapsibility, conditional independence. |
|
Number of Downloads: 261 | Number of Views: 799 |
|