### Table 3 Marginal likelihoods1

1997

"... In PAGE 11: ... The most obvious feature of Figure 2 is the distance of the posterior interquartile values from the population values of P d = 1 x () for the conventional probit model and the corresponding closeness for the mixture models. This is reflected in the marginal likelihood values in the second row of Table3 . Interquartile ranges are larger for the two-mixture model than for the conventional probit model and slightly larger yet for the three-mixture model.... In PAGE 11: ... Interquartile ranges are larger for the two-mixture model than for the conventional probit model and slightly larger yet for the three-mixture model. But the fit of the conventional probit model is so bad that its probability is very low relative to the other two (as indicated in the second row of Table3 ). The three-mixture model looks a little closer to the DGP than does the two-mixture model in the interquartile range metric, but the... ..."

### Table 2: The log marginal and posterior marginal likelihoods for the factor analysis model of the language data in Fuller (1987). ( * maximum marginal likelihood )

1997

"... In PAGE 13: ... The eigenvalues of the data covariance matrix are diag? = (207:339; 15:079; 13:848; 10:185; 10:026; 9:400; 8:038; 6:572); and the cumulated percentages are pj = Pj i=1 i P8 i=1 i ; j = 1; : : : ; 8; the results are given by Table 1. Table2 lists the (ordinary) marginal like- lihood and the posterior marginal likelihood (see Appendix D) for di erent number of factors. Two factors are chosen by the maximum marginal likeli- hood criterion for the model with and without outliers.... ..."

Cited by 1

### Table 2: Log marginal likelihoods for various models (relative to saturated)

"... In PAGE 5: ... This is con rmed by calculating, using Laplace apos;s method, the marginal likelihoods for a variety of models. These are displayed in Table2 , with their corresponding dimensions. U1 0 20 40 60 80 100 120 U2 0 20 40 60 80 100 120 U3 0 50 100 150 200 U4 0 50 100 150 200 250... In PAGE 6: ...Table 2: Log marginal likelihoods for various models (relative to saturated) The model where is constrained to lie in U1 has a marginal likelihood over 20 000 times as great as any other model. This model incorporates both Quasi-independence and Symmetry and is denoted in Table2 by QI+Symmetry. The posterior density for the projection distance M1 for this model is not very di erent from the distribution displayed in Figure 1, although it has a smaller variance, as would be expected when other terms have been removed from the model.... ..."

### Table 3 Marginalized likelihoods for Meat model, Multivariate regression

1994

Cited by 16

### Table 4 Marginalized likelihoods for Meat model, Reduced rank regression

1994

Cited by 16

### Table 3 Marginalized likelihoods for Meat model, Multivariate regression

1994

Cited by 16

### Table 4 Marginalized likelihoods for Meat model, Reduced rank regression

1994

Cited by 16

### Table 1: Comparison of approximated log marginal likelihoods for the GLMM with identity link

2006

"... In PAGE 15: ... For more details, see Sinharay and Stern (2001) and the references therein. Table1 shows the results which show that all methods work basically the same. Sensitivity of the results to the prior speci cation was assessed by repeating the analyses with the following di erent hyperparameters: (a) priors with variance /2; (b) priors with variance 2; (c) priors with moderately di erent means.... ..."

Cited by 4

### Table 1. The shape of the marginal likelihood/posterior with uniform prior

2003

Cited by 2

### Table 1: The log marginal likelihood of AR model, the log of average marginal likelihood (6) and of AR outlier model for Swiss GNP and consumption from 1970 Q1 to 1993 Q2 with fractional prior and Bayes factor (28) ( * maximum marginal likelihood )

"... In PAGE 22: ...U1 = Y ? X ^ B; ^ B = (X0X)?1X0Y; ^ U2 = Y ? X ^ B ? D ^ A; ( ^ B; ^ A)0 = ( ~ X0 ~ X)?1 ~ X0Y; and ~ X = (X : D ): 4 Examples Table1 analyses the (real) Swiss GNP and (real) Swiss Consumption for the period 1970 Q1 to 1993 Q2 for fractional prior where the Bayes factor is calculated by formula (53). For both time series the marginal likelihood selects the lag order p = 2 and in both cases the outlier model is selected also for lag 2.... In PAGE 22: ... Lag order 5 is found for GNP and lag 2 for consumption and the unit root model (non-stationarity) is accepted. Table 3 is an extended analysis for the outlier model in Table1 . Now we include a break point model as an alternative and we discover that the break point model with outliers ts the data best.... ..."