泊松回归

• 定义：当因变量是单位时间某事件的发生次数时就应该考虑用泊松回归，比如想研究的因变量是过去一个月病人发生脑卒中的次数，就考虑用泊松回归。

1. 分析代码：

data(Affairs, package="AER")
summary(Affairs)
table(Affairs$affairs)

# fit full model
fit.full <- glm(affairs ~ gender + age + yearsmarried + children + 
                  religiousness + education + occupation +rating,
                data=Affairs,family=poisson())
summary(fit.full)

2. OR值计算

# interpret coefficients
coef(fit.reduced)
# OR值 回归系数的反对数
exp(coef(fit.reduced))
# OR值的置信区间
exp(confint(fit.reduced))
# OR值和置信区间
exp(cbind(OR = coef(fit.reduced), confint(fit.reduced)))