15分钟进行单因素分析，多因素鉴定绘制森林图及建立随机森林模型（

首先利用compareGroups包绘制基线信息表及进行单因素逻辑回归，然后利用forestmodel进行多因素逻辑回归并绘制森林图，最后应用randomForest包进行随机森林的构建并进行验证，验证指标主要根据C指数。

library(ggpubr)
library(survminer)
library(survival)
library(randomForest)
library(compareGroups)
library(rms) 
library(forestmodel)
library(pROC)

df <- read.csv("rawdata.csv",header = T)
df$status<-factor(df$status,labels=c("alive","dead"))#生存状态因子化变为二分变量并赋值
df$sex<-factor(df$sex,labels=c("male","female"))#性别因子化变为二分变量并赋值
df$obstruct<-factor(df$obstruct,labels=c("no","yes"))#梗阻情况因子化变为二分变量并赋值
df$differ <- as.factor(df$differ)
set.seed(2021)
ind <- sample(2, nrow(df), 
              replace = T, 
              prob = c(0.7, 0.3))#有放回回抽样，建模组70%，验证组30%
devData <- df[ind == 1,]  # 建模组
vadData <- df[ind == 2,]  # 验证组

mixdata<-cbind(ind,df) #最终数据集

tab1<- descrTable(ind~.-ind,method = 1, data = mixdata)
export2word(tab1, file='table1.docx')

tab2_uni<- descrTable(status~ .,show.ratio =TRUE,
           data = df,method=c(time=2),
           show.p.trend=TRUE) 
export2word(tab2_uni, file='tab2_uni.docx')

##多因素回归结果用森林图表示
glmmodel <- glm(status ~ rx + sex + obstruct + differ, binomial(link="logit"), data = devData)

forest_model(glmmodel)

####建立逻辑回归模型并计算C指数
str(df$status)
pr <- predict(glmmodel,type=c("response"))
roccurve <- roc(df$status ~ pr)
plot.roc(roccurve,xlim = c(1,0),ylim=c(0,1))