问题
In a pandas dataframe created like this:
import pandas as pd
import numpy as np
df = pd.DataFrame(np.random.randint(10, size=(6, 6)),
columns=['c' + str(i) for i in range(6)],
index=["r" + str(i) for i in range(6)])
which could look as follows:
c0 c1 c2 c3 c4 c5
r0 2 7 3 3 2 8
r1 6 9 6 7 9 1
r2 4 0 9 8 4 2
r3 9 0 4 3 5 4
r4 7 6 8 8 0 8
r5 0 6 1 8 2 2
I can easily select certain rows and/or a range of columns using .loc:
print df.loc[['r1', 'r5'], 'c1':'c4']
That would return:
c1 c2 c3 c4
r1 9 6 7 9
r5 6 1 8 2
So, particular rows/columns I can select in a list, a range of rows/columns using a colon.
How would one do this in R? Here and here one always has to specify the desired range of columns by their index but one cannot - or at least I did not find it - access those by name. To give an example:
df <- data.frame(c1=1:6, c2=2:7, c3=3:8, c4=4:9, c5=5:10, c6=6:11)
rownames(df) <- c('r1', 'r2', 'r3', 'r4', 'r5', 'r6')
The command
df[c('r1', 'r5'),'c1':'c4']
does not work and throws an error. The only thing that worked for me is
df[c('r1', 'r5'), 1:4]
which returns
c1 c2 c3 c4
r1 1 2 3 4
r5 5 6 7 8
But how would I select the columns by their name and not by their index (which might be important when I drop certain columns throughout the analysis)? In this particular case I could of course use grep but how about columns that have arbitrary names?
So I don't want to use
df[c('r1', 'r5'),c('c1','c2', 'c3', 'c4')]
but an actual slice.
EDIT:
A follow-up question can be found here.
回答1:
It looks like you can accomplish this with a subset:
> df <- data.frame(c1=1:6, c2=2:7, c3=3:8, c4=4:9, c5=5:10, c6=6:11)
> rownames(df) <- c('r1', 'r2', 'r3', 'r4', 'r5', 'r6')
> subset(df, select=c1:c4)
c1 c2 c3 c4
r1 1 2 3 4
r2 2 3 4 5
r3 3 4 5 6
r4 4 5 6 7
r5 5 6 7 8
r6 6 7 8 9
> subset(df, select=c1:c2)
c1 c2
r1 1 2
r2 2 3
r3 3 4
r4 4 5
r5 5 6
r6 6 7
If you want to subset by row name range, this hack would do:
> gRI <- function(df, rName) {which(match(rNames, rName) == 1)}
> df[gRI(df,"r2"):gRI(df,"r4"),]
c1 c2 c3 c4 c5 c6
r2 2 3 4 5 6 7
r3 3 4 5 6 7 8
r4 4 5 6 7 8 9
回答2:
An alternative approach to subset if you don't mind to work with data.table would be:
data.table::setDT(df)
df[1:3, c2:c4, with=F]
c2 c3 c4
1: 2 3 4
2: 3 4 5
3: 4 5 6
This still does not solve the problem of subsetting row range though.
回答3:
Adding onto @evan058's answer:
subset(df[rownames(df) %in% c("r3", "r4", "r5"),], select=c1:c4)
c1 c2 c3 c4
r3 3 4 5 6
r4 4 5 6 7
r5 5 6 7 8
But note, the : operator will probably not work here; you will have to write out the name of each row you want to include explicitly. It might be easier to group by a particular value of one of your other columns or to create an index column as @evan058 mentioned in comments.
回答4:
A solution using dplyr package but you need to specify the row you want to select before hand
rowName2Match <- c("r1", "r5")
df1 <- df %>%
select(matches("2"):matches("4")) %>%
add_rownames() %>%
mutate(idRow = match(rowname, rowName2Match)) %>%
slice(which(!is.na(idRow))) %>%
select(-idRow)
df1
> df1
Source: local data frame [2 x 4]
rowname c2 c3 c4
<chr> <int> <int> <int>
1 r1 2 3 4
2 r5 6 7 8
回答5:
This seems way too easy so perhaps I'm doing something wrong.
df <- data.frame(c1=1:6, c2=2:7, c3=3:8, c4=4:9, c5=5:10, c6=6:11,
row.names=c('r1', 'r2', 'r3', 'r4', 'r5', 'r6'))
df[c('r1','r2'),c('c1','c2')]
c1 c2
r1 1 2
r2 2 3
来源：https://stackoverflow.com/questions/37714152/how-to-select-range-of-columns-in-a-dataframe-based-on-their-name-and-not-their