Using a Bayesian model for bankruptcy prediction : a comparative approach

The purpose of this study is to examine the impact of the choice of cut-off points, sampling procedures, and the business cycle on the accuracy of bankruptcy prediction models. Misclassification can result in erroneous predictions leading to prohibitive costs to firms, investors and the economy. To test the impact of the choice of cut-off points and sampling procedures, three bankruptcy prediction models are assessed- Bayesian, Hazard and Mixed Logit. A salient feature of the study is that the analysis includes both parametric and nonparametric bankruptcy prediction models. A sample of firms from Lynn M. LoPucki Bankruptcy Research Database in the U. S. was used to evaluate the relative performance of the three models. The choice of a cut-off point and sampling procedures were found to affect the rankings of the various models. In general, the results indicate that the empirical cut-off point estimated from the training sample resulted in the lowest misclassification costs for all three models. Although the Hazard and Mixed Logit models resulted in lower costs of misclassification in the randomly selected samples, the Mixed Logit model did not perform as well across varying business-cycles. In general, the Hazard model has the highest predictive power. However, the higher predictive power of the Bayesian model, when the ratio of the cost of Type I errors to the cost of Type II errors is high, is relatively consistent across all sampling methods. Such an advantage of the Bayesian model may make it more attractive in the current economic environment. This study extends recent research comparing the performance of bankruptcy prediction models by identifying under what conditions a model performs better. It also allays a range of user groups, including auditors, shareholders, employees, suppliers, rating agencies, and creditors' concerns with respect to assessing failure risk.

Vitamin D status and latent tuberculosis infection : a preliminary study in a group of healthy Mexican agricultural workers

Vitamin D metabolites are important in the regulation of bone and calcium homeostasis, but also have a more ubiquitous role in the regulation of cell differentiation and immune function. Severely low circulating 25-dihydroxyvitamin D [25(OH)D] concentrations have been associated with the onset of active tuberculosis (TB) in immigrant populations, although the association with latent TB infection (LTBI) has not received much attention. A previous study identified the prevalence of LTBI among a sample of Mexican migrant workers enrolled in Canada's Seasonal Agricultural Workers Program (SA WP) in the Niagara Region of Ontario. The aim of the present study was to determine the vitamin D status of the same sample, and identify if a relationship existed with LTBI. Studies of vitamin D deficiency and active TB are most commonly carried out among immigrant populations to non-endemic regions, in which reactivation of LTBI has occurred. Currently, there is limited knowledge of the association between vitamin D deficiency and LTBI. Entry into Canada ensured that these individuals did not have active TB, and L TBI status was established previously by an interferon-gamma release assay (IGRA) (QuantiFERON-TB Gold In-Tube®, Cellestis Ltd., Australia). Awareness of vitamin D status may enable individuals at risk of deficiency to improve their nutritional health, and those with LTBI to be aware of this risk factor for disease. Prevalence of vitamin D insufficiency among the Mexican migrant workers was determined from serum samples collected in the summer of 2007 as part of the cross sectional LTBI study. Samples were measured for concentrations of the main circulating vitamin D metabolite, 25(OH)D, with a widely used 1251 250HD RIA (DiaSorin Inc.®, Stillwater, MN), and were categorized as deficient «37.5 nmoI/L), insufficient (>37.5 nmollL, < 80 nmol/L) or sufficient (2::80 nmoI/L). Fisher's exact tests and t tests were used to determine if vitamin D status (sufficiency or insufficiency) or 25(OH)D concentrations significantly differed by sex or age categories. Predictors of vitamin D insufficiency and 25(OH)D concentrations were taken from questionnaires carried out during the previous study, and analyzed in the present study using multiple regression prediction models. Fisher's exact test and t test was used to determine if vitamin D status or 25(OH)D concentration differed by LTBI status. Strength of the relationship between interferongamma (IFN-y) concentration (released by peripheral T cells in response to TB antigens) and 25(OH)D concentration was analyzed using a Spearman correlation. Out of 87 participants included in the study (78% male; mean age 38 years), 14 were identified as LTBI positive but none had any signs or symptoms of TB reactivation. Only 30% of the participants were vitamin D sufficient, whereas 68% were insufficient and 2% were deficient. Significant independent predictors of lower 25(OH)D concentrations were sex, number of years enrolled in the SA WP and length of stay in Canada. No significant differences were found between 25(OH)D concentrations and LTBI status. There was a significant moderate correlation between IFN-y and 25(OH)D concentrations ofLTBI-positive individuals. The majority of participants presented with Vitamin D insufficiency but none were severely deficient, indicating that 25(OH)D concentrations do not decrease dramatically in populations who temporarily reside in Canada but go back to their countries of origin during the Canadian winter. This study did not find a statistical relationship between low levels of vitamin D and LTBI which suggests that in the presence of overall good health, lower than ideal levels of 2S(OH)D, may still be exerting a protective immunological effect against LTBI reactivation. The challenge remains to determine a critical 2S(OH)D concentration at which reactivation is more likely to occur.

Predicting the Risk of Lung Cancer in Never-smokers

Despite being considered a disease of smokers, approximately 10-15% of lung cancer cases occur in never-smokers. Lung cancer risk prediction models have demonstrated excellent ability to discriminate cases from non-cases, and have been shown to be more efficient at selecting individuals for future screening than current criteria. Existing models have primarily been developed in populations of smokers, thus there was a need to develop an accurate model in never-smokers. This study focused on developing and validating a model using never-smokers from the Prostate, Lung, Colorectal, and Ovarian Cancer Screening Trial. Cox regression analysis, with six-year follow-up, was used for model building. Predictors included: age, body mass index, education level, personal history of cancer, family history of lung cancer, previous chest X-ray, and secondhand smoke exposure. This model achieved fair discrimination (optimism corrected c-statistic = 0.6645) and good calibration. This represents an improvement on existing neversmoker models, but is not suitable for individual-level risk prediction.

Optimal and Robust Designs of Step-stress Accelerated Life Testing Experiments for Proportional Hazards Models

Accelerated life testing (ALT) is widely used to obtain reliability information about a product within a limited time frame. The Cox s proportional hazards (PH) model is often utilized for reliability prediction. My master thesis research focuses on designing accelerated life testing experiments for reliability estimation. We consider multiple step-stress ALT plans with censoring. The optimal stress levels and times of changing the stress levels are investigated. We discuss the optimal designs under three optimality criteria. They are D-, A- and Q-optimal designs. We note that the classical designs are optimal only if the model assumed is correct. Due to the nature of prediction made from ALT experimental data, attained under the stress levels higher than the normal condition, extrapolation is encountered. In such case, the assumed model cannot be tested. Therefore, for possible imprecision in the assumed PH model, the method of construction for robust designs is also explored.