Statistical appraisal in solving some medical problems / Lim Fong Peng

• Main Author: Lim, Fong Peng
• Format: Thesis
• Published: 2016
• Subjects: Q Science (General); R Medicine (General); T Technology (General)
• Online Access: http://studentsrepo.um.edu.my/6772/1/fong_peng.pdf; http://studentsrepo.um.edu.my/6772/
• Institution: Universiti Malaya

Interest in some medical problems has raised the need for the development of appropriate statistical techniques in order to provide reliable solutions. We look at two local medical scenarios which are of current interest; firstly, identifying the optimal number of lymph nodes removed for maximizing the survival and adequate nodal staging of local breast cancer patients, and secondly, studying the outlier detection in cross-over design for kinesiology study. In this thesis, we will discuss alternative and new methods to provide the solution to the scenarios above. For the breast cancer study, we investigate the influence of the number of lymph nodes removed (LNR) on survival of breast cancer patients using Chi-square test of independence and Wilcoxon test. We proceed to find the best-fitted logistic and Cox’s regression models using forward selection and Bayesian model averaging procedures. The models are then used to assess the prognostic values of independent factors for survival at all thresholds of the number of LNR. For both types of regression models, we use not only the Wald statistic but also present the use of the Akaike Information Criterion to determine the optimal number of LNR which results in maximum differential in survival of the breast cancer patients. Similar procedure will be extended to the case of finding the dependence of number of LNR to the adequate nodal staging of the patients. For the kinesiology study, we employ both non-Bayesian and Bayesian framework to detect outliers in a 2 × 2 cross-over design. We consider the mixed model with different factors representing subject, period, treatment and carry-over effects. In non-Bayesian framework, we consider the classical studentized residual and provide a studentized residual using median absolute deviation to identify possible outlying subjects. The performances of both procedures in detecting subject outliers are compared via simulation. On the other hand, in Bayesian framework, we assume that the random subject effect and the errors are normal distributed. However, the outlying subjects come from normal distribution with different variance. Due to the complexity of the resulting joint posterior distribution, we obtain the information on the posterior distribution from samples by using Markov Chain Monte Carlo method. We use two real data sets, the Malaysian Breast Cancer data and kinesiology data, obtained from the University of Malaya Medical Centre (UMMC). This study is able to provide solutions to the problems which are very beneficial to the local medical practitioners. The findings are very important as guidelines in the surgical management of breast cancer patients and in the usage of kinesiotapes in sports.