Question

I have a simple problem, honestly I have tried to find an answer. I really did.

I have a bunch of .csv files that have been imported into R data frames

I would like to take a specific column (with a common name) from each data frame, merge it into a single data frame with the name of the data frame as the column name, and produce a boxplot using each column.

The columns are not of the same length and frequently contain NA’s.

Example: Data frames (in which the first row is the header)

Data frame name Tom:

col1    col2 col3 col4
name1   33   44   55
name2   33   NA   55
name3   33   34   55
name4   33   24   55

Data frame name Bob:

col1    col2 col3 col4
name5   33   74   55
name6   33   NA   55
name7   33   32   55

Data frame name Stu:

col1    col2 col3 col4
name8   33   44   55
name9   33   11   55
name10  33   34   55
name11  33   24   55
name12  33   32   55
name13  33   24   5
name14  33   34   55
name15  33   24   5

Desired result

Tom  Bob  Stu
44   74   44
NA   NA   11 
34   32   34
24        24
          32
          24
          34
          24

So, taking “col3” (the column name is shared) from each data frame, and produce a new data frame of only the col3 data, each column to be named as the name of the data frame it came from…followed by producing a side-by-side boxplot of Tom, Bob and Stu (but I can probably work that out). It’s OK to have NA’s in the empty spaces in the desired result above.

Answer 1

Here is a basic approach where I create a new combined data frame using rbind, after adding an identifier column to each of your 3 data frames. Note that you can also create a boxplot without first creating a single data frame.

Tom = read.table(header=TRUE,
text="col1    col2 col3 col4
name1   33   44   55
name2   33   NA   55
name3   33   34   55
name4   33   24   55")

Bob = read.table(header=TRUE,
text="col1    col2 col3 col4
name5   33   74   55
name6   33   NA   55
name7   33   32   55")

Stu = read.table(header=TRUE,
text="col1    col2 col3 col4
name8   33   44   55
name9   33   11   55
name10  33   34   55
name11  33   24   55
name12  33   32   55
name13  33   24   5
name14  33   34   55
name15  33   24   5")

# Add a new person identifier column to each data frame.
Tom$person = "Tom"
Bob$person = "Bob"
Stu$person = "Stu"

# Combine 3 data frames by row.
dat = rbind(Tom, Bob, Stu)

dat
#      col1 col2 col3 col4 person
# 1   name1   33   44   55    tom
# 2   name2   33   NA   55    tom
# 3   name3   33   34   55    tom
# 4   name4   33   24   55    tom
# 5   name5   33   74   55    bob
# 6   name6   33   NA   55    bob
# 7   name7   33   32   55    bob
# 8   name8   33   44   55    stu
# 9   name9   33   11   55    stu
# 10 name10   33   34   55    stu
# 11 name11   33   24   55    stu
# 12 name12   33   32   55    stu
# 13 name13   33   24    5    stu
# 14 name14   33   34   55    stu
# 15 name15   33   24    5    stu


boxplot(col3 ~ person, data=dat)

# This would also work, without rearranging the data:
boxplot(Tom[, "col3"], Bob[, "col3"], Stu[, "col3"])

# Save to pdf file.
pdf("boxplot_1.pdf", height=5, width=5)
boxplot(col3 ~ person, data=dat, main="Boxplot of three samples.", ylab="col3")
dev.off()