How to apply a common baseline across multiple words

[stonekate]

Hey Bruno,

Ages ago I mentioned I wanted to apply the baseline for one word to multiple words in a sentence, and you said send you what I tried... finally I got around to it. Here I've tried to compute the mean amplitude across the baseline for the word I want and subtract it from all the other words, but it doesn't really work. Can you see what I'm doing wrong? I put the corresponding data and script (common_baseline_example.R) in Dropbox/kate_eeg_data

TIA

load(tidyverse)
load(eeguana)


dat <- readRDS("test.rds")


# extract the baseline of the word whose baseline we want to apply to all other words
# i.e. the word before the target word in 4 conditions:
bl_segs <- dat %>%
  eeg_segment(lead(.description %in% c("S200", "S210", "S220", "S230")), 
              .lim = c(-.2, 0), .unit = "s")

# compute the average amplitude for each of these baseline segments
bls <- bl_segs %>% eeg_group_by(.id) %>% eeg_summarise(across_ch(mean))

# subtract the baseline from all segments corresponding with that .id.
# It gets the first row correct but no others, and just takes the first row
# of bls and applies it, regardless of .id (e.g. first .id of bls is 1 and of
# dat_seg is 46)
dat2 <- dat %>% eeg_group_by(.id) %>% eeg_mutate(across_ch(~.x - bls$.signal$.x))

[stonekate]

Ok I maybe came up with a (very clunky) solution (posting here in case it's useful or someone one day can suggest something better):

My issue is that I'm creating a separate eeg_lst with the baseline average amplitudes at the first word in a sentence which should then be applied to every word in that same sentence. If there are 20 sentences, then this yields an eeg_lst with one baseline value per sentence, .id values 1:20.

But each of the words in the 20 sentences in the main eeg_lst gets its own .id, so there are .ids 1:nrow(eeg_lst). Thus, .id 2 in the baseline eeg_lst is for the first word of the second sentence, but .id 2 in the main eeg_lst is the second word of the first sentence: The issue was therefore how to match the .id in the baseline eeg_lst with the relevant sentences in the main eeg_lst.

To do this:

• First I relabeled the BrainVision triggers (.description) in the events table with numeric values so that each trigger in one sentence had a trigger that---when rounded down---had the same value. E.g. sentence 1 had separate words labelled 200, 200.01, 200.02, 200.03 etc., sentence 2, 210, 210.01, 210.02, 210.03 etc.
• Then I set an id_counter with initial value 1
• Then I set previous_trigger_value with initial value equal to the first trigger in the events table
• Then I wrote a for loop looping over .id in the main eeg_lst where every time the corresponding rounded-down sentence trigger value changed, id_counter went up by one
• The baseline eeg_lst was filtered by id_counter
• The main eeg_lst was filtered by the current loop iteration value
• The baseline subset was subtracted from the subset of the main eeg_lst
• The now-baselined subset of the main eeg_lst was bound to a new eeg_lst
• previous_trigger_value was updated with the current trigger value

So the loop extracts each individual word from the main eeg_lst, subtracts the relevant baseline, then adds it to a new eeg_lst, word-by-word. Clunky but it seems to work.