count true false sum values by time thresholds in r

I have a student dataset that includes responses to questions as right or wrong. There is also a time variable in seconds. I would like to create a time flag to record number of correct and incorrect responses by 1 minute 2 minute and 3 minute thresholds. Here is a sample dataset.

df <- data.frame(id = c(1,2,3,4,5),
gender = c("m","f","m","f","m"),
age = c(11,12,12,13,14),
i1 = c(1,0,NA,1,0),
i2 = c(0,1,0,"1]",1),
i3 = c("1]",1,"1]",0,"0]"),
i4 = c(0,"0]",1,1,0),
i5 = c(1,1,NA,"0]","1]"),
i6 = c(0,0,"0]",1,1),
i7 = c(1,"1]",1,0,0),
i8 = c(0,0,0,"1]","1]"),
i9 = c(1,1,1,0,NA),
time = c(115,138,148,195, 225))
> df
id gender age i1 i2 i3 i4   i5 i6 i7 i8 i9 time
1  1      m  11  1  0 1]  0    1  0  1  0  1  115
2  2      f  12  0  1  1 0]    1  0 1]  0  1  138
3  3      m  12 NA  0 1]  1 <NA> 0]  1  0  1  148
4  4      f  13  1 1]  0  1   0]  1  0 1]  0  195
5  5      m  14  0  1 0]  0   1]  1  0 1] NA  225

The minute thresholds are represented by a ] sign at the right side of the score.

For example for the id = 3, the 1-minute threshold is at item i3 , the 2-minute threshold is at item i6. Each student might have different time thresholds.

I need to create flagging variables to count number of correct and incorrect responses by the 1-min 2-min and 3-min thresholds.

How can I achieve the desired dataset as below.

> df1
id gender age i1 i2 i3 i4   i5 i6 i7 i8 i9 time one_true one_false two_true two_false three_true three_false
1  1      m  11  1  0 1]  0    1  0  1  0  1  115        2         1       NA        NA         NA          NA
2  2      f  12  0  1  1 0]    1  0 1]  0  1  138        2         2        4         3         NA          NA
3  3      m  12 NA  0 1]  1 <NA> 0]  1  0  1  148        1         1        2         2         NA          NA
4  4      f  13  1 1]  0  1   0]  1  0 1]  0  195        2         0        3         2          5           3
5  5      m  14  0  1 0]  0   1]  1  0 1] NA  225        1         2        2         3          4           4

Here's a dplyr pipe that produces what you want.

I'm optionally using xfun::n2w to convert 1 to "one", etc. If you can accept strict numbers, then you don't need this.

library(dplyr)
library(tidyr) # pivot_*
# library(xfun) # n2w, convert numbers to words
df %>%
  select(-gender, -age, -time) %>%
  mutate(across(-id, as.character)) %>%
  pivot_longer(-id) %>%
  arrange(id, name) %>%
  mutate(
    grp = 1L + cumsum(grepl("]", lag(value), fixed=TRUE)),
    grp = xfun::n2w(grp),
    num = if_else(gsub("[^0-9]", "", value) == "1", "true", "false"),
    .by = id) %>%
  filter(
    any(grepl("]", value, fixed = TRUE)), 
    .by = c(id, grp)) %>%
  count(id, grp, num) %>%
  filter(!is.na(num)) %>%
  mutate(n = cumsum(n), .by = c(id, num)) %>%
  pivot_wider(
    id, names_sep = "_",
    names_from = c("grp", "num"), values_from = "n"
  ) %>%
  left_join(df, ., by = "id")
#   id gender age i1 i2 i3 i4   i5 i6 i7 i8 i9 time one_false one_true two_false two_true three_false three_true
# 1  1      m  11  1  0 1]  0    1  0  1  0  1  115         1        2        NA       NA          NA         NA
# 2  2      f  12  0  1  1 0]    1  0 1]  0  1  138         2        2         3        4          NA         NA
# 3  3      m  12 NA  0 1]  1 <NA> 0]  1  0  1  148         1        1         2        2          NA         NA
# 4  4      f  13  1 1]  0  1   0]  1  0 1]  0  195        NA        2         3        5           1          4
# 5  5      m  14  0  1 0]  0   1]  1  0 1] NA  225         2        1         4        4           3          3

This is using .by=, so required dplyr_1.1.0 or newer.

Most of the work is done on a temporary frame that returns id and the intended summary columns one_false and beyond.

• mutate(across(-id, as.character)) because pivot_longer requires compatible classes, and your sample data here has some columns int, some chr.
• pivot_longer reshapes from "wide" to three columns: id, name ("i1", "i2", ...), and value ("1", "0", "1]", ...)
• mutate(grp...) condition the groups such that everything up until the first occurrence of ] is in group "one", "two", etc; the use of xfun:n2w to go from 2 to "two" is purely aesthetic, you can do without if you can accept column names starting with numbers, ala 1_true (assuming you mutate num as well);
• mutate(num...) converts your 1s and 0s to "true" and "false"; this is mostly aesthetic, included to match your intended output; if you would prefer just 0 and 1, then you'd still need to remove the ] in order to count things correctly
• filter(any(...)) removes rows that do not end in a value with ]
• count counts (weird) by the groups
• mutate(n=cumsum(n)) by different grouping ensures your one_true and two_true are cumulative
• pivot_wider converts multiple rows into columns, undoing our first effort
• we bring that summary back to the original df with left_join