In dealing with data we often encounter should be variable is the two classification.docVIP

 PAGE \* MERGEFORMAT 9 In dealing with data we often encounter should be variable is the two classification [Keywords:] in dealing with data In dealing with data, we often encounter the dependent variable is the two categories, multiple categories, such as: disease or not the disease, success and failure, treatment outcome of cure, improvement, invalid and so on, assuming that the probability of occurrence P, n-observations are independent under the conditions generally used Logistic regression analysis of data, the model variables can be continuous and categorical variables can also be applied very extensive. SAS software provides a Logistic, genmod, nlmixed, catmod such processes to analyze the data, this study discusses how to use different processes to analyze the two multi-category classification and Logistic regression model. A dependent variable for the two categories Logistic Regression Example 1: To study of gender and the different treatments of a disease curable or not, the impact of the data in Table 1. Table 1 Gender and treatment of a disease effects (abbreviated) Program: data aa; input sex $ treat $ y count @ @; cards; ma 1 78 ma 0 28 mb 1 101 mb 0 11 ... ... fc 0 6; Method 1: PROC GENMOD data = aa descending; class sex treat; freq count; model response = sex treat / dist = binomial link = logit; run; / * dist = binomial and link = logit fit two categories logistic regression model * / Method 2: PROC LOGISTIC data = aa descending; class sex treat / param = ref ref = last; freq count; model response = sex treat; run; Method 3: proc catmod order = data; weight count; model y = sex treat; run; in order to explain the intuitive nature, we usually have to be interested in routed to the final value of the variable (in ascending order), the fit of the event (event) probability of occurrence, cases of a selection of choice items in descending, fitting the dependent variable y = 1 the probability that the fitting ordered value of a corresponding