Statistical Analysis and Modeling of Brain Tumor Data: Histology and Regional Effects

• Main Author: Pokhrel, Keshav Prasad
• Format: Others
• Published: Scholar Commons 2013
• Subjects: brain tumor; functional; mortality; probability density function; quantile; Epidemiology; Statistics and Probability
• Online Access: http://scholarcommons.usf.edu/etd/4746; http://scholarcommons.usf.edu/cgi/viewcontent.cgi?article=5943&context=etd

Comprehensive statistical models for non-normally distributed cancerous tumor sizes are of prime importance in epidemiological studies, whereas a long term forecasting models can facilitate in reducing complications and uncertainties of medical progress. The statistical forecasting models are critical for a better understanding of the disease and supply appropriate treatments. In addition such a model can be used for the allocations of budgets, planning, control and evaluations of ongoing efforts of prevention and early detection of the diseases. In the present study, we investigate the effects of age, demography, and race on primary brain tumor sizes using quantile regression methods to obtain a better understanding of the malignant brain tumor sizes. The study reveals that the effects of risk factors together with the probability distributions of the malignant brain tumor sizes, and plays significant role in understanding the rate of change of tumor sizes. The data that our analysis and modeling is based on was obtained from Surveillance Epidemiology and End Results (SEER) program of the United States. We also analyze the discretely observed brain cancer mortality rates using functional data analysis models, a novel approach in modeling time series data, to obtain more accurate and relevant forecast of the mortality rates for the US. We relate the cancer registries, race, age, and gender to age-adjusted brain cancer mortality rates and compare the variations of these rates during the period of the study that data was collected. Finally, in the present study we have developed effective statistical model for heterogenous and high dimensional data that forecast the hazard rates with high degree of accuracy, that will be very helpful to address subject health problems at present and in the future.