还有一个疑问:
Call:
glm(formula = flag ~ X1wt + X1tt + X1zf + X1zp + X2wt + X2tt +
X2zf + X2zp, family = binomial(), data = datasample)
Deviance Residuals:
Min 1Q Median 3Q Max
-1.938 -1.252 0.774 0.905 1.495
Coefficients:
Estimate Std. Error z value Pr(>|z|)
(Intercept) 8971.4616 6768.6303 1.325 0.1850
X1wt -2.1824 1.9992 -1.092 0.2750
X1tt -146.9639 113.6175 -1.293 0.1958
X1zf 0.2125 0.3947 0.538 0.5903
X1zp 1728.6228 1318.8870 1.311 0.1900
X2wt 7.6280 3.9075 1.952 0.0509 .
X2tt 59.5489 103.7609 0.574 0.5660
X2zf 0.5778 0.4304 1.342 0.1795
X2zp -769.1400 1195.9249 -0.643 0.5201
---
Signif. codes: 0 ‘***’ 0.001 ‘**’ 0.01 ‘*’ 0.05 ‘.’ 0.1 ‘ ’ 1
(Dispersion parameter for binomial family taken to be 1)
Null deviance: 812.58 on 632 degrees of freedom
Residual deviance: 780.10 on 624 degrees of freedom
AIC: 798.1
Number of Fisher Scoring iterations: 4
上面是我把多个变量当作预测变量去做logistic回归,结果每个都不显著
但用单个变量当作预测变量去做logistic回归,显示却是显著的,如下用X1wt 做预测变量
Call:
glm(formula = flag ~ X1wt, family = binomial(), data = datasample)
Deviance Residuals:
Min 1Q Median 3Q Max
-1.7745 -1.2838 0.8103 0.8906 1.3475
Coefficients:
Estimate Std. Error z value Pr(>|z|)
(Intercept) -146.626 34.212 -4.286 1.82e-05 ***
X1wt 1.119 0.260 4.305 1.67e-05 ***
---
Signif. codes: 0 ‘***’ 0.001 ‘**’ 0.01 ‘*’ 0.05 ‘.’ 0.1 ‘ ’ 1
(Dispersion parameter for binomial family taken to be 1)
Null deviance: 812.58 on 632 degrees of freedom
Residual deviance: 793.77 on 631 degrees of freedom
AIC: 797.77
Number of Fisher Scoring iterations: 4
请问这种情况怎么解释?
|