Abstract: This paper continues the introduction to minimum encoding inductive inference
given by Oliver and Hand. This series of papers was written with the objective of providing
an introduction to this area for statisticians. We describe the message length estimates ...

... A comparison of MML and MDL can be found in Baxter and Oliver 3]. ... In the example discussed
in Section 4.1 (models consisting of normal curves), the model space can be represented as
a two dimensional space with 2 (?1; 1), and 2 (0; 1). The Original MML Method partitions ...

Abstract This paper continues the introduction to minimum encoding inductive inference
given by Oliver and Hand. This series of papers was written with the objective of providing
an introduction to this area for statisticians. We describe the message length estimates ...

... We tested the effectiveness of the methods for selecting the number of components in a mixture
model by repeat- edly generating data from artificial mixture ... We table the predicted number of
components for each method in Section 3, namely MML, AIC, PC, MDL and ICOMP. ...

... similar approach to that taken in [18] in that a simplifica- tion of the Hessian evaluation is ... from
those of all other components, hence, the determinant of the Hessian of the complete model will
be ... we obtain, not- ing that there are only K - 1 free parameters, so the differ- entiation is ...

... 2. SOME PROBLEMS WITH MDL Earlier comparisons between earlier versions of MDL
and MML are given in [3, 7] and the accompanying discussion in that 1987 issue of JR Statist.
Soc.; and also in [8]. 2.1. Partitioning models into 'model classes' ...

... we believe it is clear that MML has performed quite well, particularly in comparison with TETRAD ...
data, and use that data as input to both TETRAD II and our MML induction program ... if a causal
induction program is working perfectly, it should reproduce ex- actly the model used to ...

... In this paper, we shall not consider the comparison of arbitrary models, but consider a more ...
mi>D</mml:mi><mml:mo stretchy="true">-</mml:mo></mml:mover></mml:mrow></mml:math>
D ... For example, Banerjee and others (2004) make extensive use of DIC for model selection. ...

... Nominal Response Model: A Comparison ... ESTIMATING NR MODEL PARAMETERS: MML VS.
MCMC 345 ... It is interesting to note that the cells for which MML worked better also tended to be
the cells where parameters were generally poorly estimated by both methods. ...