Adaptive prior weighting in generalized regression

Power priors for replication studies

The ongoing replication crisis in science has increased interest in the methodology of replication studies. We propose a novel Bayesian analysis approach using power priors: The likelihood of the original study's data is raised to the power of α , and then used as the prior distribution in the analysis of the replication data. Posterior distribution and Bayes factor hypothesis tests related to the power parameter α quantify the degree of compatibility between the original and replication study. Inferences for other parameters, such as effect sizes, dynamically borrow information from the original study. The degree of borrowing depends on the conflict between the two studies. The practical value of the approach is illustrated on data from three replication studies, and the connection to hierarchical modeling approaches explored. We generalize the known connection between normal power priors and normal hierarchical models for fixed parameters and show that normal power prior inferences with a beta prior on the power parameter α align with normal hierarchical model inferences using a generalized beta prior on the relative heterogeneity variance I 2 . The connection illustrates that power prior modeling is unnatural from the perspective of hierarchical modeling since it corresponds to specifying priors on a relative rather than an absolute heterogeneity scale.

Bias in odds ratios from logistic regression methods with sparse data sets

BACKGROUND

Logistic regression models are widely used to evaluate the association between a binary outcome and a set of covariates. However, when there are few study participants at the outcome and covariate levels, the models lead to bias of the odds ratio (OR) estimated using the maximum likelihood (ML) method. This bias is known as sparse data bias, and the estimated OR can yield impossibly large values because of data sparsity. However, this bias has been ignored in most epidemiological studies.

METHODS

We review several methods for reducing sparse data bias in logistic regression. The primary aim is to evaluate the Bayesian methods in comparison with the classical methods, such as the ML, Firth's, and exact methods using a simulation study. We also apply these methods to a real data set.

RESULTS

Our simulation results indicate that the bias of the OR from the ML, Firth's, and exact methods is considerable. Furthermore, the Bayesian methods with hyper-g prior modeling of the prior covariance matrix for regression coefficients reduced the bias under the null hypothesis, whereas the Bayesian methods with log F-type priors reduced the bias under the alternative hypothesis.

CONCLUSION

The Bayesian methods using log F-type priors and hyper-g prior are superior to the ML, Firth's, and exact methods when fitting logistic models to sparse data sets. The choice of a preferable method depends on the null and alternative hypothesis. Sensitivity analysis is important to understand the robustness of the results in sparse data analysis.

Prior Elicitation for Use in Clinical Trial Design and Analysis: A Literature Review

Bayesian inference is increasingly popular in clinical trial design and analysis. The subjective knowledge derived from an expert elicitation procedure may be useful to define a prior probability distribution when no or limited data is available. This work aims to investigate the state-of-the-art Bayesian prior elicitation methods with a focus on clinical trial research. A literature search on the Current Index to Statistics (CIS), PubMed, and Web of Science (WOS) databases, considering “prior elicitation” as a search string, was run on 1 November 2020. Summary statistics and trend of publications over time were reported. Finally, a Latent Dirichlet Allocation (LDA) model was developed to recognise latent topics in the pertinent papers retrieved. A total of 460 documents pertinent to the Bayesian prior elicitation were identified. Of these, 213 (45.4%) were published in the “Probability and Statistics” area. A total of 42 articles pertain to clinical trial and the majority of them (81%) reports parametric techniques as elicitation method. The last decade has seen an increased interest in prior elicitation and the gap between theory and application getting narrower and narrower. Given the promising flexibility of non-parametric approaches to the experts’ elicitation, more efforts are needed to ensure their diffusion also in applied settings.

The assessment of intrinsic credibility and a new argument for p < 0.005

The concept of intrinsic credibility has been recently introduced to check the credibility of 'out of the blue' findings without any prior support. A significant result is deemed intrinsically credible if it is in conflict with a sceptical prior derived from the very same data that would make the effect just non-significant. In this paper, I propose to use Bayesian prior-predictive tail probabilities to assess intrinsic credibility. For the standard 5% significance level, this leads to a new p-value threshold that is remarkably close to the recently proposed p < 0.005 standard. I also introduce the credibility ratio, the ratio of the upper to the lower limit (or vice versa) of a confidence interval for a significant effect size. I show that the credibility ratio has to be smaller than 5.8 such that a significant finding is also intrinsically credible. Finally, a p-value for intrinsic credibility is proposed that is a simple function of the ordinary p-value and has a direct frequentist interpretation in terms of the probability of replicating an effect. An application to data from the Open Science Collaboration study on the reproducibility of psychological science suggests that intrinsic credibility of the original experiment is better suited to predict the success of a replication experiment than standard significance.

Bayesian logistic regression approaches to predict incorrect DRG assignment

Episodes of care involving similar diagnoses and treatments and requiring similar levels of resource utilisation are grouped to the same Diagnosis-Related Group (DRG). In jurisdictions which implement DRG based payment systems, DRGs are a major determinant of funding for inpatient care. Hence, service providers often dedicate auditing staff to the task of checking that episodes have been coded to the correct DRG. The use of statistical models to estimate an episode's probability of DRG error can significantly improve the efficiency of clinical coding audits. This study implements Bayesian logistic regression models with weakly informative prior distributions to estimate the likelihood that episodes require a DRG revision, comparing these models with each other and to classical maximum likelihood estimates. All Bayesian approaches had more stable model parameters than maximum likelihood. The best performing Bayesian model improved overall classification per- formance by 6% compared to maximum likelihood, with a 34% gain compared to random classification, respectively. We found that the original DRG, coder and the day of coding all have a significant effect on the likelihood of DRG error. Use of Bayesian approaches has improved model parameter stability and classification accuracy. This method has already lead to improved audit efficiency in an operational capacity.

Bias away from the null due to miscounted outcomes? A case study on the TORCH trial

Count outcomes occur in virtually all disciplines, such as medicine, epidemiology or biology, but they often contain error. Recently, it has been shown that self-reported numbers of exacerbations of Chronic Obstructive Pulmonary Disease patients can be considerably miscounted. Motivated by this result, we reanalysed data from the Towards a Revolution in Chronic Obstructive Pulmonary Disease Health trial, a large randomized controlled trial with the self-reported number of exacerbations of Chronic Obstructive Pulmonary Disease patients as outcome. To adjust for miscounting error in the response of Poisson and (zero-inflated) negative binomial models, we introduce novel, general methodology. The key idea is to formulate a zero-inflated negative binomial model to capture the error mechanism. This parametric approach automatically circumvents drawbacks of previously suggested methodology that treats miscounted outcomes in the misclassification framework. Prior information for the response error model parameters was elicited from validation data of an external study and adaptively weighted to account for potential prior-data conflict. The results of the Bayesian hierarchical modelling approach indicated that the treatment effect has been overestimated in the original study. However, closer inspection revealed that this unexpected result was an artefact of an unaccounted time dependency of the treatment effect.