R中具有多个虚拟编码预测变量的线性回归模型的箱线图

Question

I have the following linear model:

model <- lm(var01 ~ a0 + a1 + a2 + a3 + a4 + a5,NT)

Where var01 is a intervall-scaled variable from 0-100 and a0-a5 are dummy coded (0, 1) variables. The summary(model) gives this:

Residuals:
    Min      1Q  Median      3Q     Max 
-75.951 -13.469  -7.239  18.795  80.531 

Coefficients:
            Estimate Std. Error t value Pr(>|t|)    
(Intercept)  59.6015     8.7076   6.845 5.48e-10 ***
a01         -46.1329     8.6302  -5.345 5.37e-07 ***
a11          -0.8744     9.0549  -0.097   0.9233    
a21          22.0408     9.1278   2.415   0.0175 *  
a31           9.5488     9.9284   0.962   0.3384    
a41          14.9227     7.6762   1.944   0.0546 .  
a51          -8.1222    11.8530  -0.685   0.4947    
---
Signif. codes:  0 ‘***’ 0.001 ‘**’ 0.01 ‘*’ 0.05 ‘.’ 0.1 ‘ ’ 1

Residual standard error: 32.13 on 104 degrees of freedom
Multiple R-squared:  0.4393,    Adjusted R-squared:  0.407 
F-statistic: 13.58 on 6 and 104 DF,  p-value: 2.486e-11

I would like to create a boxplot where a0-a5 are displayed next to each other, but only with a0==1,a1==1, etc.

So I tried:

ggplot(NT, aes(factor(a0), var01)) +
  geom_boxplot() +
  geom_smooth(method = "lm", se=FALSE, color="black", aes(group=1))

But this shows the boxplots for a0 == 0 and a0 == 1 next to each other. So two questions: How do I get R only to show a0 == 1? And furthermore all four other predictors a1-a5 next to a0 (but also limited to a1-a4 == 1) in the same graphic?

Help is very appreciated. Thanks :)

Update: Sample data

id  category_a  var01   a0  a1  a2  a3  a4  a5
3   1;5          100    0   1   0   0   0   1
4   1;5            0    0   1   0   0   0   1
5   0             21    1   0   0   0   0   0
6   1;2;4        100    0   1   1   0   1   0
9   1;2           68    0   1   1   0   0   0

So a0-a5 are dummy codings of multi-category variable "category_a".

Answer 1

It's a question of data-reshaping. ggplot works best if each data-point you're interested in is one row in a dataframe (long format).

library(ggplot2)
library(reshape2)
#generate data
set.seed(1)
n=1000
NT <- data.frame(id=1:n,
                   var01=rnorm(n),
                   a0=rbinom(n,1,0.2),
                   a1=rbinom(n,1,0.2),
                   a2=rbinom(n,1,0.2),
                   a3=rbinom(n,1,0.2),
                   a4=rbinom(n,1,0.2),
                   a5=rbinom(n,1,0.2))

#do some data-reshaping before plotting
#ggplot needs each data-point on one line
#so transform to long
plotdata <- melt(NT,id.vars=c("id","var01"),variable.name="a")

Now it's very easy to plot everything:

#plot everything using interaction
p1 <- ggplot(plotdata, aes(x=interaction(a,value), y=var01)) +
  geom_boxplot()
p1

Or a selection:

p2 <- ggplot(plotdata[plotdata$value==1,], 
             aes(x=a, y=var01)) +
  geom_boxplot()

p2