R ggplot 四次多项式的误差线

发布于 2025-01-18 00:21:31 字数 1067 浏览 1 评论 0原文

我正在尝试使用R的统计学习介绍来重现此图的右侧（图7.1，p267）：

这是我正在使用的代码，而曲线似乎在点上，错误条不会在我的图表中的高值飞行，但相反会变得更窄：


library(tidyverse)
library(ISLR2)
library(broom)

wage <- Wage %>%
  as_tibble() %>%
  select(age, wage) %>%
  mutate(high_inc = wage > 250)

# 7.1.B Graph (ISLR page 267)
glm(high_inc ~ poly(age, 4), data = wage, family = binomial) %>%
  broom::augment(type.predict = 'response') %>%
  cbind(wage %>% select(age)) %>%
  as_tibble() %>%
  mutate(lower = .fitted - 1.96 * .se.fit,     
         upper = .fitted + 1.96 * .se.fit) %>% # POINTWISE STANDARD ERRORS
  ggplot(aes(age, .fitted)) +
  geom_line() +
  geom_ribbon(aes(ymin = lower, ymax = upper), color = "grey", alpha = .2)

src =“ https://i.sstatic.net/vrxp4.jpg” alt =“在此处输入图像说明”>

有人知道为什么这是/我在做什么错？

原文

I'm trying to reproduce the right side of this graph from Introduction to Statistical Learning using R (Figure 7.1, p267):

Here is the code I'm using, and while the curve seems on point, the error bars do not fly off wildly for high values in my graph, but conversely get more narrow:


library(tidyverse)
library(ISLR2)
library(broom)

wage <- Wage %>%
  as_tibble() %>%
  select(age, wage) %>%
  mutate(high_inc = wage > 250)

# 7.1.B Graph (ISLR page 267)
glm(high_inc ~ poly(age, 4), data = wage, family = binomial) %>%
  broom::augment(type.predict = 'response') %>%
  cbind(wage %>% select(age)) %>%
  as_tibble() %>%
  mutate(lower = .fitted - 1.96 * .se.fit,     
         upper = .fitted + 1.96 * .se.fit) %>% # POINTWISE STANDARD ERRORS
  ggplot(aes(age, .fitted)) +
  geom_line() +
  geom_ribbon(aes(ymin = lower, ymax = upper), color = "grey", alpha = .2)