R plot boxplot per column in dataframe and customizing parameters

Question

I have an R dataframe with 20 columns, one for each model. The lines of the dataset present the statistics for a boxplot. I want to plot a boxplot for each of those models, setting the parameters of the boxplot as the lines of the dataframe

Below is one example:

        Model 1    Model 2   ...  Model 20
min       1           5              15
q25       2           7              16
median    3           8              20
q75       4           9              21
max       5           10             22

As can be seeing, the statistics are already calculated. I just need to set them to the boxplot but I have no idea in how to do that

Does this answer your question? [Draw bloxplots in R given 25,50,75 percentiles and min and max values](https://stackoverflow.com/questions/11129432/draw-bloxplots-in-r-given-25-50-75-percentiles-and-min-and-max-values) — jpsmith, Sep 03 '22 at 23:04
Curious, how did you calculate those stats? If in R, run `boxplot` on that previous step. — Parfait, Sep 03 '22 at 23:51
@jpsmith I am having a weird behaviour. `bxp` expects a list. I pass a list and I receive an error: `Error in is.finite(z$stats) : default method not implemented for type 'list'`. I am literally using the same code as shown in question you sent and passing my df as a list to the `bxp` — Rods2292, Sep 04 '22 at 12:46

OliverHennhoefer · Answer 1 · 2022-09-04T06:54:32.090

In case you are willing to use ggplot2 you could try something like this:

Set up a fake dataset. Apparently, you need that to run ggplot() + geom_boxplot():

df <- data.frame("Model" = "Model 1")

Then you can control the single boxplot components like this:

ggplot(df, aes(x = Model,
           ymin=5,     #min
           lower=20,   #q25
           middle=25,  #median
           upper=50,   #q75
           ymax=100))+ #max
  geom_boxplot(stat="identity")

Analogous for multiple models:

df <- data.frame("Model" = c("Model 1", "Model 2"))

ggplot(df, aes(x = Model,
           ymin=c(5, 9),
           lower=c(20,46),
           middle=c(25,55),
           upper=c(50,89),
           ymax=c(100, 111)))+
  geom_boxplot(stat="identity")

score 0 · Answer 2 · answered Sep 04 '22 at 18:47

What has not been explained so far is that you need a matrix and not a data frame (since data frames are actually lists, the error refers to lists). I assume you somewhere also have the sample sizes, I rbind them here as a new row.

dat <- rbind(dat, n=c(20, 14, 60))

So all you need to do is coercing as.matrix.

bxp(list(stats=as.matrix(dat[1:5, ]), n=dat[6, ]))

Data:

dat <- read.table(header=TRUE, text='Model1    Model2   Model20
min       1           5              15
q25       2           7              16
median    3           8              20
q75       4           9              21
max       5           10             22')

R plot boxplot per column in dataframe and customizing parameters

2 Answers2