征求意见，根据 R 中的 R 平方值从线性回归模型得出绘制大变量最好使用 GGPLOT2

问题描述

我必须为我的项目构建 231 个线性回归模型。运行 231 个模型后，我只剩下 231 个 R 平方值，我必须在图中针对变量名称显示这些值。由于 231 个 R 平方值对于一个表来说太多了，我正在寻找一个绘图的想法，这样我就可以将 R 平方值显示为 y 轴，将变量名称显示为 x 轴。当我运行 dput(head(df,5)) 时，我得到了这个（这可能会让您了解我的数据）：

structure(list(Band = c(402,411,419,427,434),R.squared = c(0.044655015122032,0.852028718800355,0.818617476505653,0.825782272278991,0.860844967662728
),Adj.Rsquared = c(-0.0614944276421867,0.835587465333728,0.798463862784058,0.806424746976656,0.845383297403031),Intercept = c(0.000142126282140086,-0.00373545760470339,-0.00258909036368109,0.000626075834918527,-3.3448513588372e-05),Slope = c(-0.00108714482110104,0.393380133190131,0.443463459485279,0.503881831479685,0.480162723468755)),row.names = c(NA,5L),class = "data.frame")

请注意，我的完整数据有 231 个观测值，我想将变量带（作为因子）绘制为 x 轴，将 R 平方绘制为 y 轴。我已经在 ggplot2 中尝试过 geom_point() 但它看起来非常混乱且难以理解。有什么想法吗？

更新：当我使用@Duck 建议的代码时，我得到了这个图，用于科学演示时有点混乱。

解决方法

如果你有大量的值，你可以在轴上避开标签，这里有一个例子：

library(ggplot2)
#Code
ggplot(mdf,aes(x=factor(Band),y=R.squared))+
  geom_point()+
  scale_x_discrete(guide = guide_axis(n.dodge=2))+
  coord_flip()

输出：

使用的一些数据：

#Data
mdf <- structure(list(Band = c(402,411,419,427,434,412,421,429,437,444,422,431,439,447,454,432,441,449,457,464),R.squared = c(0.044655015122032,0.852028718800355,0.818617476505653,0.825782272278991,0.860844967662728,0.044655015122032,0.860844967662728),Adj.Rsquared = c(-0.0614944276421867,0.835587465333728,0.798463862784058,0.806424746976656,0.845383297403031,-0.0614944276421867,0.845383297403031),Intercept = c(0.000142126282140086,-0.00373545760470339,-0.00258909036368109,0.000626075834918527,-3.3448513588372e-05,0.000142126282140086,-3.3448513588372e-05
    ),Slope = c(-0.00108714482110104,0.393380133190131,0.443463459485279,0.503881831479685,0.480162723468755,-0.00108714482110104,0.480162723468755)),row.names = c(NA,-20L),class = "data.frame")

@DaveArmstrong 的建议也很有帮助（非常感谢他）：

#Code 2
ggplot(mdf,aes(x=reorder(factor(Band),R.squared,mean),y=R.squared))+
  geom_point()+
  scale_x_discrete(guide = guide_axis(n.dodge=2))+
  coord_flip()

输出：

另一种选择：

#Code 3
ggplot(mdf,y=R.squared))+
  geom_point()+
  geom_segment( aes(x=reorder(factor(Band),xend=reorder(factor(Band),y=0,yend=R.squared))+
  scale_x_discrete(guide = guide_axis(n.dodge=2))+
  coord_flip()

输出：

,

热图怎么样。

dat <- tibble(
  Band = 100:330,R.squared = runif(231,1),Adj.Rsquared = R.squared - runif(231,.1)
) %>% 
  mutate(Band = reorder(factor(Band),mean)) 

q <- c(0,quantile(dat$R.squared,c(.2,.4,.6,.8)),1)
dat <- dat %>% mutate(group=cut(R.squared,breaks=q))
levels(dat$group) <- c("Quntile 1","Quintile 2","Quintile 3","Quintile 4","Quintile 5")

ggplot(dat,aes(x=1,y = Band,fill=R.squared)) + 
  geom_tile(col="white") + 
  facet_wrap(~group,scales="free_y",nrow=1) + 
  scale_x_continuous(breaks=NULL) + 
  scale_fill_viridis_c() + 
  labs(x = "",y="",fill=expression(R^2)) + 
  theme(legend.position="top",axis.text.y = element_text(size=10))

这将是另一种选择，其中标签位于单元格内：

R2 <- matrix(dat$R.squared,ncol=21,nrow=11)
txt <- matrix(as.character(dat$Band),nrow=11)

dat2 <- tibble(
  R.squared = c(R2),Band = c(txt),row = rep(1:11,21),col = rep(21:1,each=11)
)

ggplot(dat2,aes(x=row,y=col,fill=R.squared)) + 
  geom_tile() + 
  geom_text(aes(label=Band),color="white") + 
  scale_fill_viridis_c(option="B") + 
  theme_void() + 
  theme(legend.position="bottom") + 
  labs(x = "",fill=expression(R^2))