Is there a reason RowSums(df[grep wouldn't work accurately? - r

Question

I used
df$Total.P.n <- rowSums(df[grep('p.n', names(df), ignore.case = FALSE)])
to sum count values from any column name containing p.n, but the values it produced are way off. The columns are counts of certain combinations of language types in a language corpus. I want to get a summary of all times p.n. was used within other combinations, but am struggling. It seems like perhaps it is counting other occurences like e.sp.NR in my variable names, but shouldn't ignore.case=FALSE take care of that? I've also tried tidyverse and dplyr solutions to no avail.
Here's example of df structure:
ID.
do.p.n.NP
do.p.n.SE
p.d.e.sp.SR
1510
4
6
2
1515
2
0
1
and what I need:
ID.
do.p.n.NP
do.p.n.SE
p.d.e.sp.SR
Total.P.n
1510
4
6
2
10
1515
2
0
1
2

Answer 1

Update after update(new column names) of OP:
The code is like:
df$Total.P.n <- rowSums(df[grep('p.n', names(df), ignore.case = FALSE)])
df$p.d.e.sp.SR <- rowSums(df[,2:3]!=0)
ID. do.p.n.NP do.p.n.SE. p.d.e.sp.SR Total.P.n
1 1510 4 6 2 10
2 1515 2 0 1 2
First answer:
The argument pattern you are searching for e.g. p.n does not exist in df. Therefore I think you mean pn: Then your code works as expectect:
df$Total.P.n <- rowSums(df[grep('pn', names(df), ignore.case = FALSE)])
ID. do.pn.NP do.pn.SE. p.d.e.sp.SR Total.P.n
1 1510 4 6 0 10
2 1515 2 0 1 2

Answer 2

If we can use dplyr, I would suggest using a tidy-select function / selection helper like matches. And please mind that your regex is likely wrong. If we need to match literal dots . , we need to escape the metacharacter with a double backslash. The appropriate regex would be n\\.p.
library(dplyr)
data
df <- tibble(`ID.` = c(1510, 1515), `do.p.n.NP` = c(4,2), `do.p.n.SE.` = c(6,0), `p.d.e.sp.SR` = c(0,1))
answer
df %>%
mutate(Total.P.n = rowSums(across(matches('p\\.n'))))
# A tibble: 2 × 5
ID. do.p.n.NP do.p.n.SE. p.d.e.sp.SR Total.P.n
<dbl> <dbl> <dbl> <dbl> <dbl>
1 1510 4 6 0 10
2 1515 2 0 1 2