估计因子变量的 logit 模型会产生错误答案

【问题标题】：Estimate logit models for factor variable produces error估计因子变量的 logit 模型会产生错误
【发布时间】：2018-04-14 12:55:02
【问题描述】：

我无法估计一个因子变量为因变量的 logit 模型。我创建了一个可重现的示例来更好地解释并显示错误消息。

## create a reproducible example that replicates the problem
set.seed(12) # reproducibility of the "randomly" generated data. 
df<-data.frame(dummy=as.factor(rep(c("yes","no"),100)), # factor encoding
               x=rnorm(n = 200,mean = 5,sd = 1)) # some predictor variable


# calculate regression with different encodings
summary(glm(formula = dummy~x,data = df)) # does not work

这种方法的错误信息是

Error in glm.fit(x = c(1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1,  : 
  NA/NaN/Inf in 'y'
    In addition: Warning messages:
1: In Ops.factor(y, mu) : ‘-’ not meaningful for factors
2: In Ops.factor(eta, offset) : ‘-’ not meaningful for factors
3: In Ops.factor(y, mu) : ‘-’ not meaningful for factors

我不太明白这条消息。数据规模（因子）有什么问题，还是我如何应用该函数的问题？任何帮助将非常感激。

【问题讨论】：

标签： r char logistic-regression

【解决方案1】：

添加family="binomial" 以指定这是一个逻辑回归并且有效：

> fit <- (glm(formula = dummy~x, data = df, family="binomial")) 
> summary(fit)

Call:
glm(formula = dummy ~ x, family = "binomial", data = df)

Deviance Residuals: 
     Min        1Q    Median        3Q       Max  
-1.18456  -1.17736  -0.00041   1.17736   1.18342  

Coefficients:
             Estimate Std. Error z value Pr(>|z|)
(Intercept)  0.028747   0.734674   0.039    0.969
x           -0.005782   0.145003  -0.040    0.968

(Dispersion parameter for binomial family taken to be 1)

    Null deviance: 277.26  on 199  degrees of freedom
Residual deviance: 277.26  on 198  degrees of freedom
AIC: 281.26

Number of Fisher Scoring iterations: 3

【讨论】：