# 第一题
sample(c("正面","反面"),
       50,
       replace = T,
       prob=c(0.55,0.45))

# 第二题
rm(list=ls())
library(tidyverse)
data("ToothGrowth")
str(ToothGrowth)
colnames(ToothGrowth)
df <- ToothGrowth |> 
  mutate(len2=ifelse(len>18.8,"阳性","阴性"))
df$len2 <- factor(df$len2,
                  levels = c("阴性","阳性"),
                  labels = c("阴性","阳性"))
# logistic
fit <- glm(len2~supp+dose,
    family = binomial(link = "logit"),
    data=df)
temp <- summary(fit)
temp <- temp$coefficients

OR <- exp(coef(fit))
OR_interval <- exp(confint(fit))
OR <- cbind(OR,OR_interval)
result <- cbind(temp,OR)
getwd()
write.csv(result,"logistic.csv",row.names = T)


# 3. *,
fit1 <- glm(len2~supp+dose+supp:dose, #写法1
           family = binomial(link = "logit"),
           data=df)
fit2 <- glm(len2~supp*dose, #写法2
            family = binomial(link = "logit"),
            data=df)

fit3 <- glm(len2~(supp+dose)^2, #写法3
            family = binomial(link = "logit"),
            data=df)

# 交互效应的解读
summary(fit3)
# 1.看主效应：看主效应方向；主效应的方向代表这个向量对结局的影响；
# 2.看交互效应的方向：交互效应的方向代表了另外一个变量对该变量对结局影响的削弱或者增强。
# 3.一般只做两个变量的交互，没必要做更高阶的交互。
# 4.交互效应有没有必要放入模型？看P值；看AIC。
# anova(fit,fit3)
AIC(fit,fit3)