2

我很难理解为什么我的数据没有显示置信区间。当我在另一个数据集上重现我的代码时，代码似乎工作正常。例如，在mtcars

代码是

mtols = mtcars %>% group_by(am) %>% do(lm0 = lm(disp ~ mpg*gear + vs, data=.)) %>% 
               augment(., lm0) %>% 
               mutate(ymin=.fitted-1.96*.se.fit, ymax=.fitted+1.96*.se.fit)

生成绘图

mtols %>% ggplot(aes(mpg, .fitted) ) + 
  geom_smooth(data = mtols, aes(mpg, .fitted, group = gear, colour = gear, fill= gear), method="lm") +
  theme_minimal() + facet_grid(~am)

我得到了置信区间。

但是，这不适用于我的数据。有人可以帮我弄清楚这里出了什么问题吗？我会很感激。

我OLS用

dt = new %>% group_by(day) %>% do(lm0 = lm(y ~ year*class, data=.)) %>% augment(., lm0) %>% 
  mutate(ymin=.fitted-1.96*.se.fit, ymax=.fitted+1.96*.se.fit) 

dt$year = as.numeric(as.character(dt$year))

情节，（这是一个少数情况的例子，但结果与整个数据集相同）

dt %>% ggplot(aes(year, .fitted) ) + 
  geom_smooth(data = dt, aes(year, .fitted, group = class, colour = class, fill= class), method="lm") + 
  theme_bw() + facet_grid(~day)

CI没有显示。

在此处输入图像描述

任何线索我在这里做错了什么？

奇怪的是，当我不使用facet_grid这里时，CI工作完美

在此处输入图像描述

dt %>% ggplot(aes(year, .fitted) ) + 
  geom_smooth(data = dt, aes(year, .fitted, group = class, colour = class, fill= class), method="lm") + 
  theme_bw()

我的数据样本

library(broom) 
library(dplyr)
library(ggplot2)

new = structure(list(id = structure(c(844084L, 114510L, 14070410L, 
942483L, 13190105L, 421369L, 301384L, 251789L, 11011210L, 11280408L, 
278575L, 310410L, 16260105L, 11110815L, 18260101L, 14260501L, 
10580L, 15090210L, 19140410L, 13230615L, 246511L, 20040812L, 
14260114L, 287623L, 16090620L, 20131007L, 835743L, 453390L, 395808L, 
363617L), label = "Household identifier", class = c("labelled", 
"integer")), year = structure(c(1L, 1L, 2L, 1L, 2L, 1L, 1L, 1L, 
2L, 2L, 1L, 1L, 2L, 2L, 2L, 2L, 1L, 2L, 2L, 2L, 1L, 2L, 2L, 1L, 
2L, 2L, 1L, 1L, 1L, 1L), .Label = c("2000", "2015"), class = "factor"), 
day = c("Weekend", "Weekend", "Weekend", "Weekdays", "Weekdays", 
"Weekend", "Weekdays", "Weekend", "Weekend", "Weekdays", 
"Weekend", "Weekdays", "Weekdays", "Weekend", "Weekend", 
"Weekdays", "Weekdays", "Weekend", "Weekdays", "Weekdays", 
"Weekdays", "Weekend", "Weekend", "Weekend", "Weekend", "Weekend", 
"Weekend", "Weekdays", "Weekdays", "Weekdays"), class = structure(c(1L, 
1L, 2L, 2L, 1L, 2L, 2L, 4L, 2L, 2L, 3L, 2L, 1L, 4L, 1L, 3L, 
2L, 3L, 2L, 4L, 2L, 1L, 3L, 2L, 1L, 4L, 3L, 2L, 4L, 1L), .Label = c("Higher Managerial", 
"Lower Managerial", "Intermediate", "Manual and Routine"), class = "factor"), 
y = c(270, 730, 180, 0, 0, 290, 90, 650, 510, 0, 10, 200, 
200, 180, 0, 0, 140, 260, 110, 740, 260, 0, 390, 610, 0, 
0, 500, 0, 10, 170)), class = "data.frame", row.names = c(NA, 
-30L), .Names = c("id", "year", "day", "class", "y"))
4

1 回答 1

1

正在绘制置信区间。我们看不到它们，因为每个只有两个独特的点day。

dt2 <- dt %>% filter(class == "Higher Managerial")
plot(.fitted ~ year, data=subset(dt2, day=="Weekend"))

在此处输入图像描述

我们看到没有刻面的区间的原因是因为当有四个点时，区间更宽。

在此处输入图像描述

当我们不按方面突破时，有足够的点数可以使信心有一定的范围。但是两点的置信区间没有范围。

confint(lm(.fitted ~ year, data=subset(dt2, day=="Weekdays")))
#                     2.5 %      97.5 %
#   (Intercept) 9503.333333 9503.333333
# year            -4.666667   -4.666667

编辑

这是一个版本，我们使用ymin和ymax最初计算，并用 绘制它geom_ribbon。

dt %>% ggplot(aes(year, .fitted,group = class, colour = class, fill= class)) + 
  geom_line() +
  geom_ribbon(aes(ymin=ymin, ymax=ymax), alpha=0.2) + 
  theme_bw() + facet_grid(~day)

在此处输入图像描述

于 2016-11-15T16:27:35.183 回答