问题描述
我只想使用 Effects 包绘制多项式回归的一个类别。
示例：
library(nnet)
m1 <- multinom(Species ~ Sepal.Width * Petal.Width + Sepal.Length,data = iris)
plt <- effect("Sepal.Width * Petal.Width",m1,x.var = "Sepal.Width")
plot(plt,x.var = "Sepal.Width",lattice = list(strip = list(factor.names = F)),confint = list(style="auto",col = "black"),axes = list(grid = T,x = list(rug = F)))
这给了我以下内容： https://i.stack.imgur.com/fdOw5.png
有什么方法可以只绘制，例如 Versicolor 类？
已经尝试使用 layout()，但我只能绘制 Setosa 和 Versicolor：
plot(plt,lattice = list(strip = list(factor.names = F),layout = c(5,2,1)),x = list(rug = F)))
如果我将 layout = c(5,1) 更改为 layout = c(5,1,1)，它只会显示 Setosa。
谢谢！
解决方法
我发现的最简单的方法是使用 ggplot2。
首先，我必须将我的 object（从 effect()）转换为数据框，然后从那里开始：
示例：
m1 <- multinom(Species ~ Sepal.Width * Petal.Width + Sepal.Length,data = iris)
plt <- effect("Sepal.Width * Petal.Width",m1)
plt_df <- data.frame(plt) # creating a data frame
并使用 ggplot2 绘图：
ggplot(data = plt_df,aes(x = Sepal.Width,y = prob.versicolor,ymin = L.prob.versicolor,# Lower CI
ymax = U.prob.versicolor)) + # Upper CI
geom_ribbon(alpha = 0.2) + # For CI
geom_line() +
facet_wrap(~Petal.Width,ncol = 5) +
# Few things to custoimze
scale_y_continuous("Predicted Probabilities Y",# axes on top / right sides
sec.axis = dup_axis(),breaks = c(0,0.25,0.5,.75,1),labels = c("0%","25%","50%","75%","100%")) +
scale_x_continuous("Sepal Width") +
theme_bw() +
theme(plot.title = element_text(hjust = 0.5),strip.placement = "outside",## If you want to point axis ticks inwards:
### (2.75pt is the default axis tick length here)
# axis.ticks.length = unit(-2.75,"pt"),## do not show top / right axis labels:
# axis.text.x.top = element_blank(),## For Secondary Axis:
#axis.text.y.right = element_blank(),## As Above,Don't show axis titles:
# axis.title.x.top = element_blank(),axis.title.y.right = element_blank(),# Change the distance between y axis title and ticks:
axis.title.y = element_text(margin = margin(
t = 0,r = 20,b = 0,l = 0)),panel.background = element_blank())
