A Two-Latent-Class Model for Smoking Cessation Data with Informative Dropouts

Non ignorable missing data is a common problem in longitudinal studies. Latent class models are attractive for simplifying the modeling of missing data when the data are subject to either a monotone or intermittent missing data pattern. In our study, we propose a new two-latent-class model for categorical data with informative dropouts, dividing the observed data into two latent classes; one class in which the outcomes are deterministic and a second one in which the outcomes can be modeled using logistic regression. In the model, the latent classes connect the longitudinal responses and the missingness process under the assumption of conditional independence. Parameters are estimated by the method of maximum likelihood estimation based on the above assumptions and the tetrachoric correlation between responses within the same subject. We compare the proposed method with the shared parameter model and the weighted GEE model using the areas under the ROC curves in the simulations and the application to the smoking cessation data set. The simulation results indicate that the proposed two-latent-class model performs well under different missing procedures. The application results show that our proposed method is better than the shared parameter model and the weighted GEE model.     

A Dynamic Latent Model for Poverty Measurement

The main purpose of this paper is the longitudinal analysis of the poverty phenomenon. By interpreting poverty as a latent variable, we are able to resort to the statistical methodology developed for latent structure analysis. In particular, we propose to use the mixture latent Markov model which allows us to achieve two goals: (i) a time-invariant classification of households into homogenous groups, representing different levels of poverty; (ii) the dynamic analysis of the poverty phenomenon which highlights the distinction between transitory and permanent poverty situations. Furthermore, we exploit the flexibility provided by the model in order to achieve the measurement of poverty in a multidisciplinary framework, using several socio-economic indicators as covariates and identifying the main relevant factors which influence permanent and transitory poverty. The analysis of the longitudinal data of the Survey on Households Income and Wealth of the Bank of Italy provides the identification of two groups of households which are characterized by different dynamic features. Moreover, the inclusion of socio-economic covariates such as level of education, employment status, geographical area and residence size of the household head shows a direct association with permanent poverty.     

Estimating Transition Probabilities for Ignorable Intermittent Missing Data in a Discrete-Time Markov Chain

This article considers a discrete-time Markov chain for modeling transition probabilities when multiple successive observations are missing at random between two observed outcomes using three methods: a na\"?ve analog of complete-case analysis using the observed one-step transitions alone, a non data-augmentation method (NL) by solving nonlinear equations, and a data-augmentation method, the Expectation-Maximization (EM) algorithm. The explicit form of the conditional log-likelihood given the observed information as required by the E step is provided, and the iterative formula in the M step is expressed in a closed form. An empirical study was performed to examine the accuracy and precision of the estimates obtained in the three methods under ignorable missing mechanisms of missing completely at random and missing at random. A dataset from the mental health arena was used for illustration. It was found that both data-augmentation and nonaugmentation methods provide accurate and precise point estimation, and that the na\"?ve method resulted in estimates of the transition probabilities with similar bias but larger MSE. The NL method and the EM algorithm in general provide similar results whereas the latter provides conditional expected row margins leading to smaller standard errors.     

Re-randomization tests as sensitivity analyses to confirm immunological noninferiority of an investigational vaccine: Case study

Here we present as case study how re-randomization tests were performed in two randomized, controlled clinical trials as sensitivity analyses, as recommended by the United States Food and Drug Administration in the context of adaptive randomization. This was done to confirm primary conclusions on immunological noninferiority of an investigational new fully liquid presentation of a quadrivalent cross-reacting material conjugate meningococcal vaccine (MenACWY-CRM), over its licensed lyophilized/liquid presentation. In two phase 2b studies (Study #1: NCT03652610; Study #2: NCT03433482), noninferiority of the fully liquid presentation of MenACWY-CRM to the licensed presentation was assessed and demonstrated for immune responses against meningococcal serogroup A (MenA), the only vaccine component modified from lyophilized to liquid in the new presentation. The original vaccine assignment algorithm, with a minimization procedure accounting for center or center within age strata, was used to re-randomize participants belonging to the fully liquid and licensed vaccine groups while keeping antibody responses, covariates and entry order as observed. Test statistics under re-randomization were generated according to the ANCOVA model used in the primary analysis. To confirm immunological noninferiority following re-randomization, the corresponding p-values had to be <0.025. For both studies and all primary objective evaluations, the re-randomization p-values were well below 0.025 (0.0004 for Study #1; 0.0001 for the two co-primary endpoints in Study #2). Re-randomization tests performed to comply with a regulatory request confirmed the primary conclusions of immunological noninferiority for the MenA of the fully liquid compared to the licensed vaccine presentation.     

Modelling Incomplete Spells of Unemployment and Employment

ABSTRACT: The procedure used to analyse a data set which includes only censored or incomplete spells is examined in this paper. First of all, the distributions of incomplete spell durations are analysed without explanatory variables (such as age, gender, and so on), assuming that unobserved completed spells have a Weibull distribution. The relationships between the mean of the incomplete spells and the mean of the completed spells are reported for first-job seekers, unemployed, employed and self-employed workers. Given that the unobserved completed spells are Weibull-distributed, the unobserved heterogeneity is introduced on the scale parameter of the Weibull. The heterogeneity, considered as a variable, is analysed for a binomial or Weibull distribution. As the beginning of a spell is a retrospective datum, the recall errors are modelled including the heaping effect. Using some proportional hazards models, the methodology to study the influences of explanatory variables on spell distributions is then described, once again including both the heterogeneity and the heaping effect. On this basis, the lengths of on-going spells of unemployment for first-job seekers and unemployed workers are modelled, as well as the current job tenures of employed and self-employed workers.     

Modeling the Audit Opinions Issued to Bankrupt Companies: A Two‐stage Empirical Analysis*

Many observers are dissatisfied with the accounting profession's ability to warn the public of upcoming bankruptcy filings. Since regulators and users tend to treat an unmodified audit opinion as a “clean bill of health,” they do not expect the business to fail in the near future. Research has shown that more often than not, auditors end up letting users down when it comes to predicting bankruptcy filings with audit opinions. Although auditors assert they are not responsible for predicting future events, it is very clear that their opinion decision is evaluated, at least in part, based on events that occur after the audit report date. The interesting and logical next step is to find out how companies exit bankruptcy. Do they liquidate or reorganize? Successful reorganization may, in the end, exonerate auditors and preserve their role as an early warning device. The opinion prediction model developed in the paper introduces a new bankruptcy resolution variable that proxies for the auditor's prognosis of the ultimate disposition of the soon‐to‐be‐bankrupt company. Using a sample of bankruptcy filings between 1982 and 1992, we find that auditors do not seem to be able to predict filings or resolution. Our tests of bankruptcy resolution support what auditors have been arguing for years: that they are not clairvoyant with respect to a client's future.     

Evaluating and Refining Expert Critiquing Systems: A Methodology*

Expert critiquing systems are a rapidly growing class of intelligent decision support systems that apply artificial intelligence techniques to criticize a user's proposed solution to a problem. Critic programs now exist in the medical, engineering, programming, knowledge acquisition, word processing, and other domains. Critic refinement is a nontrivial activity that, even when done well, consumes a sizable fraction of the complete effort to build and deploy a critiquing system. To ease that effort, it is important to adopt a rigorous approach that allows one to reproducibly measure the degree of success of the current critic version and to predict which refinements will improve the critic further. The current article presents one such approach with actual case studies that illustrate its usage and elaborate selected aspects of the refinement process.