Question

我知道可以用AICcmodavg获得预测值（原始尺度〜概率）和SE的固定效果，但是我没有成功地尝试...有人可以帮助我吗？提前致谢

library(lme4)
(gm1 <- glmer(cbind(incidence, size - incidence) ~ period + (1 | herd),
              data = cbpp, family = binomial))
fixef(gm1)

library("AICcmodavg")

predictSE(gm1, 
          newdata=as.data.frame(period=c("period1","period2","period3","period4")), 
          type="response", 
          se.fit=TRUE, 
          level=0, 
          print.matrix=F)

Answer 1

最好是阅读levels(cbpp$period)，而不是as.data.frame()，而是data.frame()

levels(cbpp$period)
# [1] "1" "2" "3" "4"

predictSE(gm1, 
          newdata = data.frame(period=c("1", "2", "3", "4")),
          type = "response", 
          se.fit = TRUE, 
          level = 0, 
          print.matrix = F)

的 [编辑]

一种查找错误原因的简单方法
fit <- ...(..., data = df) predictSE(fit, newdata = df) predictSE(fit, newdata = ...) # If 1st predictSE() doesn't run, it means the model causes error. # If 1st runs but 2nd doesn't, it means it is due to newdata.
如果你的模型有两个因素;
newd <- expand.grid(name1 = levels(df$name1), name2 = levels(df$name2)) predictSE(fit, newdata = newd) # pred <- predictSE(fit, newdata = newd) # cbind(newd, pred) # help to interpret

glmer AICcmodavg的预测值

1 个答案: