TESTING Colon CANCER SURVIVAL Years by A NEW Method

In previous clinical evaluation experiments, the basic hypotheses in both paired t-test for two samples and ANOVA (analysis of variance) for multi-samples are that sample distribution is normal, and variance exhibits homoscedasticity. The apparent advantage of using non-parametric tests is that it is not necessary to examine whether or not the grouped data is normally distributed. However, when a group has heteroscedasticity, there is no test available to analyze it directly. Therefore, this study develops a new method to solve above problems. The principle is that fourier-transformed parameters are asymptotically normal, followed by a Monte Carlo simulation to compare the survival month of colon cancer patients in different groups (stages, sites etc.). In the evaluation of pre-clinical studies and clinical trials, the Kruskal-Wallis test, is frequently used as a non-parametric method for testing if the tendency for observations, from at least one sample, is larger or smaller than in the others. This effectively tests the hypothesis of equal distributions of three or more independent samples. Compared to the Kruskal-Wallis test, this new test is more accurate in experiments with sample heterogeneity. Finally, this study uses actual data (from Chang Gung Memorial Hospital, Linkou, Taiwan) from colon rectal surgery in a medical center in northern Taiwan to analyze the survival rates of different tumor location groups.