Australian & New Zealand Journal of Statistics最新文献

Semi-supervised learning is being extensively applied to estimate classifiers from training data in which not all the labels of the feature vectors are available. We present gmmsslm, an R package for estimating the Bayes' classifier from such partially classified data in the case where the feature vector has a multivariate Gaussian (normal) distribution in each of the pre-defined classes. Our package implements a recently proposed Gaussian mixture modelling framework that incorporates a missingness mechanism for the missing labels in which the probability of a missing label is represented via a logistic model with covariates that depend on the entropy of the feature vector. Under this framework, it has been shown that the accuracy of the Bayes' classifier formed from the Gaussian mixture model fitted to the partially classified training data can even have lower error rate than if it were estimated from the sample completely classified. This result was established in the particular case of two Gaussian classes with a common covariance matrix. Here we focus on the effective implementation of an algorithm for multiple Gaussian classes with arbitrary covariance matrices. A strategy for initialising the algorithm is discussed and illustrated. The new package is demonstrated on some real data.

We propose a new class of priors for Bayesian hypothesis testing, which we name ‘cake priors’. These priors circumvent the Jeffreys–Lindley paradox (also called Bartlett's paradox) a problem associated with the use of diffuse priors leading to nonsensical statistical inferences. Cake priors allow the use of diffuse priors (having one's cake) while achieving theoretically justified inferences (eating it too). We demonstrate this methodology for Bayesian hypotheses tests for various common scenarios. The resulting Bayesian test statistic takes the form of a penalised likelihood ratio test statistic. Under typical regularity conditions, we show that Bayesian hypothesis tests based on cake priors are Chernoff consistent, that is, achieve zero type I and II error probabilities asymptotically. We also discuss Lindley's paradox and argue that the paradox occurs with small and vanishing probability as sample size increases.

The COVID-19 pandemic caused an unprecedented excess mortality. Since 2020, many studies have focussed on the characteristics of COVID-19 patients who did not survive. From the statistical point of view, what seems to dominate is the large heterogeneity of the populations affected by COVID-19 and the extreme difficulty in identifying subpopulations who died affected by a plurality of contemporary characteristics. In this paper, we propose an extremely flexible approach based on a cluster-weighted model, which allows us to identify latent groups of patients sharing similar characteristics at the moment of hospitalisation as well as a similar mortality. We focus on one of the hardest hit areas in Italy and study the heterogeneity in the population of patients affected by COVID-19 using administrative data on hospitalisations in the first wave of the pandemic. Results highlighted that a model-based clustering approach is essential to understand the complexity of the COVID-19 patients treated by hospitals and who die during hospitalisation.

Response models used in marketing are not always constructed for later marketing optimisation, which often results in unsatisfactory results in target selection for future marketing activities. To solve this problem, we develop a new binary response model and a new marketing target selection method. The proposed model can predict multiple propensity scores per customer through customer-specific propensity score distributions, which is not possible with existing response models, filling a gap in the literature. The target selection method can determine the best propensity scores from those predicted by the proposed model and use them to select customers for further marketing activities. Our simulation results and application to real marketing data confirm that the performance of the proposed model in target selection is significantly better than that of the existing models, including some popular machine learning methods, which indicate that our method can be very useful in practice.

Changes in the distribution of income over time are identified based on an adjusted two-sample version of the Neyman smooth test by using subsampling methods to approximate the sampling distribution of the test statistic when samples are not independent of each other. A range of Monte Carlo experiments show that the approach corrects for size distortions arising from dependent samples as well as generating monotonic power functions. Applying the approach to studying the distribution of income in Australia over the business cycle and the Global Financial Crisis, the empirical results highlight the importance of higher-order moments and demonstrate that business cycles are not all alike as the relative strengths of higher-order moments vary over phases of the cycle.

Two-way layouts are common in grain industry research where it is often the case that there are one or more covariates. It is widely recognised that when estimating fixed effect parameters, one should also examine for possible extra error variance structure. An exact test for heteroscedasticity, when there is a covariate, is illustrated for a data set from frost trials in Western Australia. While the general algebra for the test is known, albeit in past literature, there are computational aspects of implementing the test for the two way when there are covariates. In this scenario the test is shown to have greater power than the industry standard, and because of its exact size, is preferable to use of the restricted maximum likelihood ratio test (REMLRT) based on the approximate asymptotic distribution in this instance. Formulation of the exact test considered here involves creation of appropriate contrasts in the experimental design. This is illustrated using specific choices of observations corresponding to an index set in the linear model for the two-way layout. Also an algorithm supplied complements the test. Comparisons of size and power then ensue. The test has natural extensions when there are unbalanced data, and more than one covariate may be present. Results can be extended to Balanced Incomplete Block Designs.