Had a quick look at your whole results. The following are my thoughts for your reference.
1. You have a small sample size (78) and a lot of predictors (x1 to x127). This is not good for logistic regresion. Based on your response (dependent) variable (0: 23 observations, 1: 55 observations), theoretically you'd better not put more than 3 predictors in your logistic model .
Based on your data, maybe loglinear model is a better choice.
If you really like to use logistic regression, pay attention to the following issues:
1. First please make sure all your questions follow the same directions (for example, 1 indicates positive, 0 indicates negative).
2. Do a univariate analysis by dependent variable, based on this, select 3-4 variables, put them in your logistic
model (just used forced entry or enter, not stepwise)
3. When you run logistic analysis, select correlation matrix.
4. In the correlatin matrix, if two variables are highly correlated (r>0.9), delete one of them, then re-run the model.
5. In your final model, the "S.E" should be very small
扫码加好友，拉您进群
全部版块
我的主页
收藏
