作者colin2758 (林。夏欧力)
看板Statistics
标题[问题] logistic regression的问题
时间Sat Jan 5 22:14:40 2008
若有一笔资料其中一个变数X，就是drug*(是离散的变数)
用minitab去做logistic regression最後p值是显着的
如下：
Logistic Regression Table
Odds 95% CI
Predictor Coef SE Coef Z P Ratio Lower Upper
Constant -8.90293 3.46412 -2.57 0.010
DRUG*
1 1.65755 0.594526 2.79 0.005 5.25 1.64 16.82
DIABETE*
1 0.619575 0.533708 1.16 0.246 1.86 0.65 5.29
CABG -0.731501 0.368515 -1.98 0.047 0.48 0.23 0.99
PYPASS 0.0464336 0.0165392 2.81 0.005 1.05 1.01 1.08
XCLAMP -0.0334880 0.0293418 -1.14 0.254 0.97 0.91 1.02
TEMP 0.190434 0.102047 1.87 0.062 1.21 0.99 1.48
LOSS 0.0006524 0.0005046 1.29 0.196 1.00 1.00 1.00
BLTIME 0.205822 0.106714 1.93 0.054 1.23 1.00 1.51
Log-Likelihood = -52.634
Test that all slopes are zero: G = 21.834, DF = 8, P-Value = 0.005
在做齐一性检定时，却发生drug和y是齐一的，ie是不显着的
如下：
Tabulated statistics: COMP*, DRUG*
Rows: COMP* Columns: DRUG*
0 1 All
0 47 39 86
54.65 45.35 100.00
82.46 68.42 75.44
41.23 34.21 75.44
43 43 86
1 10 18 28
35.71 64.29 100.00
17.54 31.58 24.56
8.77 15.79 24.56
14 14 28
All 57 57 114
50.00 50.00 100.00
100.00 100.00 100.00
50.00 50.00 100.00
57 57 114
Cell Contents: Count
% of Row
% of Column
% of Total
Expected count
Pearson Chi-Square = 3.030, DF = 1, P-Value = 0.082
Likelihood Ratio Chi-Square = 3.063, DF = 1,
P-Value = 0.080 >0.05
请问各位，什麽原因会造成这样问题？
--
※ 发信站: 批踢踢实业坊(ptt.cc)
◆ From: 163.23.231.83
1F：推 pigchang:或许可以解释成, 後者的检定敏感度比较低噜... 01/06 03:15