Pitfalls of estimating the marginal likelihood using the modified harmonic mean

Chan, Joshua C. C.; Grant, Angelia L.

doi:10.1016/j.econlet.2015.03.036

Cited by 27 publications

(6 citation statements)

References 18 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…For instance, one could estimate the marginal likelihood using the modified harmonic mean (Gelfand & Dey, ) of the conditional likelihood. However, Chan and Grant () find that this approach does not work well in practice, as the resulting estimates have substantial bias and tend to select the wrong model. Frühwirth‐Schnatter and Wagner () reach the same conclusion when Chib's method is used in conjunction with the conditional likelihood.…”

Section: Bayesian Model Comparison Criteriamentioning

confidence: 99%

“…However, recent work has shown that this approach can be extremely inaccurate. For example, Chan and Grant () find that the marginal likelihood estimates computed using the modified harmonic mean (Gelfand & Dey, ) of the conditional likelihood can have a substantial bias and tend to select the wrong model. Frühwirth‐Schnatter and Wagner () conclude the same when Chib's method (Chib, ) is used.…”

Section: Introductionmentioning

confidence: 99%

See 1 more Smart Citation

Bayesian model comparison for time‐varying parameter VARs with stochastic volatility

Chan

Eisenstat

2018

J of Applied Econometrics

153

140

View full text Add to dashboard Cite

Summary We develop importance sampling methods for computing two popular Bayesian model comparison criteria, namely, the marginal likelihood and the deviance information criterion (DIC) for time‐varying parameter vector autoregressions (TVP‐VARs), where both the regression coefficients and volatilities are drifting over time. The proposed estimators are based on the integrated likelihood, which are substantially more reliable than alternatives. Using US data, we find overwhelming support for the TVP‐VAR with stochastic volatility compared to a conventional constant coefficients VAR with homoskedastic innovations. Most of the gains, however, appear to have come from allowing for stochastic volatility rather than time variation in the VAR coefficients or contemporaneous relationships. Indeed, according to both criteria, a constant coefficients VAR with stochastic volatility outperforms the more general model with time‐varying parameters.

show abstract

Section: Bayesian Model Comparison Criteriamentioning

confidence: 99%

Section: Introductionmentioning

confidence: 99%

Bayesian model comparison for time‐varying parameter VARs with stochastic volatility

Chan

Eisenstat

2018

J of Applied Econometrics

153

140

View full text Add to dashboard Cite

show abstract

“…More precisely, on the methodological side, the literature makes out two popular Bayesian methods for TVP-VARs with stochastic volatility: Marginal Likelihood (ML), evaluating how likely the observed data are occurred within the system, and Deviance Information Criterion (DIC), trading off between model fit and model complexity. As regards ML estimates, they are usually obtained by using the harmonic mean 3 of a conditional likelihood 4 that tends to have a substantial bias selecting the wrong model (see, for instance, Chan and Grant (2015) and Frühwirth-Schnatter and Wagner (2008)). Concerning DIC procedure, the MCMC integration based on the conditional likelihood tends to associate higher probability to the most complex models (overfitting 5 ).…”

Section: Introductionmentioning

confidence: 99%

Structural Panel Bayesian VAR with Multivariate Time-Varying Volatility to Jointly Deal with Structural Changes, Policy Regime Shifts, and Endogeneity Issues

Pacifico

2021

Econometrics

View full text Add to dashboard Cite

This paper improves a standard Structural Panel Bayesian Vector Autoregression model in order to jointly deal with issues of endogeneity, because of omitted factors and unobserved heterogeneity, and volatility, because of policy regime shifts and structural changes. Bayesian methods are used to select the best model solution for examining if international spillovers come from multivariate volatility, time variation, or contemporaneous relationship. An empirical application among Central-Eastern and Western Europe economies is conducted to describe the performance of the methodology, with particular emphasis on the Great Recession and post-crisis periods. A simulated example is also addressed to highlight the performance of the estimating procedure. Findings from evidence-based forecasting are also addressed to evaluate the impact of an ongoing pandemic crisis on the global economy.

show abstract

“…The reason for using the one‐step‐ahead predictive likelihood as compared to the harmonic mean estimator as in Canova and Ciccarelli () is that recent work has shown that the latter approach can be extremely inaccurate. More precisely, Chan and Grant () show that the 14 marginal likelihood estimates computed using the (modified) harmonic mean as in Gelfand and Dey () can have a substantial finite‐sample bias and can thus lead to inaccurate model selection.…”

Section: Resultsmentioning

confidence: 99%

Assessing the Synchronicity and Nature of Australian State Business Cycles

Poon

2018

Economic Record

View full text Add to dashboard Cite

This paper assesses the synchronicity and nature of Australian state business cycles. To this end, I develop a time‐varying parameter panel Bayesian vector autoregression (BVAR) with a novel common stochastic volatility factor in the error structure, which is estimated in an efficient Markov chain Monte Carlo algorithm. The common stochastic volatility factor reveals that macroeconomic volatility was more pronounced during the Asian financial crisis than during the more recent global financial crisis. Next, the panel BVAR's common, state‐ and variable‐specific indicators capture several interesting economic facts. In particular, the fluctuations of the common indicator closely follow the trend line of the Organisation for Economic Co‐operation and Development composite leading indicators for Australia, making it a good proxy for nationwide business cycle fluctuations. Furthermore, despite significant co‐movements of Australian state and territory business cycles during times of economic contractions, the state indicators suggest that the average degree of synchronisation across the Australian state and territory cycles in the 2000s is only half of that in the 1990s. Given that aggregate macroeconomic activity is determined by cumulative activity of each of the states, the results suggests that the federal government should consider granting state governments greater autonomy in handling state‐specific cyclical fluctuations.

show abstract

Pitfalls of estimating the marginal likelihood using the modified harmonic mean

Cited by 27 publications

References 18 publications

Bayesian model comparison for time‐varying parameter VARs with stochastic volatility

Bayesian model comparison for time‐varying parameter VARs with stochastic volatility

Structural Panel Bayesian VAR with Multivariate Time-Varying Volatility to Jointly Deal with Structural Changes, Policy Regime Shifts, and Endogeneity Issues

Assessing the Synchronicity and Nature of Australian State Business Cycles

Contact Info

Product

Resources

About