Comparacion de estimadores de funciones de sobrevida mediante bootstrap.

Two samples of breast cancer patients with complete follow of up 25 years, were studied to assess the possible effect evolution of treatment of the disease over time. Several survival estimators were used, including Kaplan and Meier´s product limit estimator and median and quartiles residual life time. Cox´s proportional hazards regression model and semiparametric quantile regression are to be used to estimate the effect of covariates. These were classified into two groups: those related to the tumor and disease characterization, including size of the tumor, and differential temperature respect to the healthy breast, estimated through infrared imaging, histological characterization of the tumor; a second group included the risk factors of acquiring the disease, such as age of the patient at diagnostic, and at first child birth, lactation, menarche, menopause, family history of breast cancer, among others. a preliminary analysis that considered the degree of advance of the disease shows that patients at stage IV of the disease had completely different response, due to the presence of metastasis, which overcame any other possibly influencing factor. Hence the analysis of the effect of risk factors will be performed only for patients at stage I, II and III. Comparison of the two groups of data will consider standard ad hoc tests and confidence bands, obtained using bootstrap resampling methods. a simulation study will be conducted to assess the effect of different proportion of censored observations and the different parameterizations of the survival and censoring time distributions, (Weibull and uniform respectively).