• Journal of the Korean Statistical Society
• /
• 제23권1호
• /
• pp.63-78
• /
• 1994

For discrimination with binary variables, reformulated full and first order Bahadur model with incomplete observations are presented. This allows prior probabilities associated with multiple population to be estimated for the sample-based classification rule. The EM algorithm is adopted to provided the maximum likelihood estimates of the parameters of interest. Some experiences with the models are evaluated and discussed.

• Journal of Preventive Medicine and Public Health
• /
• 제49권1호
• /
• pp.45-52
• /
• 2016

Objectives: Previous large-scale cohort studies conducted in Korea have found a positive association between diabetes mellitus (DM) and colorectal cancer (CRC) in men only, in contrast to studies of other populations that have found significant associations in both men and women. Methods: A total of 1070 CRC cases and 2775 controls were recruited from the National Cancer Center, Korea between August 2010 and June 2013. Self-reported DM history and the duration of DM were compared between cases and controls. Odds ratios (ORs) and 95% confidence intervals (CIs) were calculated by binary and polytomous logistic regression models. Results: DM was associated with an elevated risk of CRC in both men (OR, 1.47; 95% CI, 1.13 to 1.90) and women (OR, 1.92; 95% CI, 1.24 to 2.98). This association remained when we controlled for age, body mass index, alcohol consumption, and physical activity level. In sub-site analyses, DM was associated with distal colon cancer risk in both men (multivariate OR, 2.04; 95% CI, 1.39 to 3.00) and women (multivariate ORs, 1.99; 95% CI, 1.05 to 3.79), while DM was only associated with rectal cancer risk in women (multivariate OR, 2.05; 95% CI, 1.10 to 3.82). No significant association was found between DM and proximal colon cancer risk in either men (multivariate OR, 1.45; 95% CI, 0.88 to 2.41) or women (multivariate OR, 1.79; 95% CI, 0.78 to 4.08). Conclusions: Overall, DM was associated with an increased risk of CRC in Koreans. However, potential over-estimation of the ORs should be considered due to potential biases from the case-control design.