【R语言】logistic回归（一）批量单因素logistic回归

---

前言

本文使用随机生成的数据集进行批量单因素logistic回归分析，并提取P<0.05的变量以供后续多因素logistic回归。后续会继续分享一些R语言分析代码，欢迎大家一起讨论学习。

---

构建数据，Y为因变量，其他为自变量

#0.构建数据，Y为因变量，其他为自变量
set.seed(1234)#设置随机种子，保证生成数据一致
log_data<- data.frame(Y = sample(0:1, 600, replace = T),
                      sex=sample(1:2, 600, replace = T),
                      edu=sample(1:4, 600, replace = T),
                      BMI=rnorm(600, mean = 22, sd = 3),
                      白蛋白=rnorm(600, mean = 35, sd = 6),
                      随机血糖=rnorm(600, mean = 4.75, sd = 1.2))
summary(log_data)
       Y               sex             edu            BMI            白蛋白         随机血糖    
 Min.   :0.0000   Min.   :1.000   Min.   :1.00   Min.   :13.21   Min.   :16.27   Min.   :1.036  
 1st Qu.:0.0000   1st Qu.:1.000   1st Qu.:2.00   1st Qu.:20.18   1st Qu.:30.94   1st Qu.:3.956  
 Median :1.0000   Median :1.000   Median :3.00   Median :22.17   Median :34.78   Median :4.816  
 Mean   :0.5017   Mean   :1.467   Mean   :2.57   Mean   :22.05   Mean   :35.04   Mean   :4.778  
 3rd Qu.:1.0000   3rd Qu.:2.000   3rd Qu.:4.00   3rd Qu.:23.98   3rd Qu.:39.19   3rd Qu.:5.579  
 Max.   :1.0000   Max.   :2.000   Max.   :4.00   Max.   :30.07   Max.   :54.01   Max.   :8.377 

---

数据处理，因子化、数值化

VarsC<-c("Y","sex","edu")
for(i in VarsC){
  log_data[,i] <- as.factor(log_data[,i])
}#利用循环因子化
summary(log_data)
Y       sex     edu          BMI            白蛋白         随机血糖    
 0:299   1:320   1:145   Min.   :13.21   Min.   :16.27   Min.   :1.036  
 1:301   2:280   2:142   1st Qu.:20.18   1st Qu.:30.94   1st Qu.:3.956  
                 3:139   Median :22.17   Median :34.78   Median :4.816  
                 4:174   Mean   :22.05   Mean   :35.04   Mean   :4.778  
                         3rd Qu.:23.98   3rd Qu.:39.19   3rd Qu.:5.579  
                         Max.   :30.07   Max.   :54.01   Max.   :8.377

---

准备进行分析的自变量

varsU<-names(log_data[,2:6])#自变量

批量单因素logistic回归

Result<-c()
for (i in 1:length(varsU)){
  fit<-glm(substitute(Y~x,list(x=as.name(varsU[i]))),data=log_data,family=binomial())
  fitSum<-summary(fit)
  result1<-c()
  result1<-rbind(result1,fitSum$coef)
  OR<-exp(fitSum$coef[,'Estimate'])
  result1<-data.frame(cbind(result1,cbind(OR,exp(confint(fit)))))
  result1$Characteristics<-varsU[i]   #添加变量名
  Result<-rbind(Result,result1[-1,])#[-1,],删除常数项
}

提取制表变量重命名，提取有意义的变量

Uni_log<-data.frame(Result[,c(1,4:8)]) #提取"P","OR","CIlower","CIupper"和变量名
colnames(Uni_log)[2:5]<-c("P","OR","CIlower","CIupper")#变量重命名
ExtractVar<-unique(Uni_log$Characteristics[Uni_log$"P"<0.05])#提取有意义的变量
write.csv(Uni_log,file="Uni_log.csv")#输出文档
Uni_log
             Estimate         P        OR   CIlower  CIupper Characteristics
sex2      0.094662809 0.5631116 1.0992881 0.7975801 1.515763             sex
edu2      0.154180710 0.5141640 1.1667017 0.7342831 1.856362             edu
edu3     -0.146213397 0.5389345 0.8639733 0.5412489 1.377153             edu
edu4      0.156454546 0.4869958 1.1693576 0.7523336 1.819676             edu
BMI       0.021802997 0.4450671 1.0220424 0.9664705 1.081118             BMI
白蛋白    0.004185709 0.7496789 1.0041945 0.9786903 1.030415          白蛋白
随机血糖 -0.076032436 0.2659728 0.9267861 0.8100385 1.059365        随机血糖