How do I filter out participants with a bad score?-CodePudding

I want to filter out participants with that score less than 45% correct. I use a pipe to group by participants, and then try to filter like this:

# Remove participants with > 45% or more mistakes
  dat <- dat %>%
    group_by(participant) %>%
      filter(sum(key_resp.corr) < (nrow(dat) * 0.65))

Key_resp.corr is 1 when the participant gave a correct answer. Unfortuneatly this doesn't seem to work. Anyone knows how I can make this work?

CodePudding user response：

Instead of nrow(dat), it would be n(), which returns the number of rows of the grouped data, whereas nrow(dat) returns the number of rows of the whole dataset

library(dplyr)
dat %>%
    group_by(participant) %>%
    filter(sum(key_resp.corr, na.rm = TRUE) < (n() * 0.65)) %>%
    ungroup

Or create the filter with mean

dat %>%
   group_by(participant) %>%
   filter(mean(key_resp.corr, na.rm = TRUE) < 0.65) %>%
   ungroup