Date of Thesis
Honors Thesis (Bucknell Access Only)
Bachelor of Science
Mark J. Meyer
statistics, odds ratios, multivariate, paired, binary, mcnemar, mental healthcare
This research was motivated by the work of Knutson et al., who were interested in pediatric mental health care coordination within the systems of care (SOC) framework, which includes mental healthcare, primary care, the educational system, child welfare, juvenile justice, and developmental disabilities. To assess the current state of care coordination, Knutson et al. specifically looked at the contacts made with 5 specific agencies within the SOC framework by primary care physicians and psychiatrists who treated the same patients. We propose an estimation and inference procedure for a common odds ratio for multivariate, paired binary data. Our closed-form odds ratio estimator is derived from the Cochran Mantel-Haenszel odds ratio estimator and for inference we used bootstrapped resampling 95% confidence intervals. We compared this procedure to three other methods of analyzing multivariate, paired binary data: conditional logistic regression, generalized estimating equations (GEE), and Cochran Mantel-Haenszel (CMH). We demonstrate that our Derived approach has just as good, if not better power than the other approaches, while maintaining reasonable Type I error. We also apply these approaches to the SOC dataset.
Horvath, Noelle Marie, "A McNemar's-Like Odds Ratio and Test for Multivariate Paired Binary Data" (2016). Honors Theses. 373.