search

peekbank / peekbank-data-import

A collection of R scripts that turn raw datasets into the peekds format ready for ingestion by the peekbank
2 stars 3 forks source link

[gazetriggered_2020] Timestamp Weirdness #35

Closed adriansteffan closed 8 months ago

adriansteffan commented 8 months ago

The point of disambiguation is sometimes very varied, even in the "original" condition of the study (where the onset was dependent on the infant looking at the screen for a certain amount of time). Some participants appear to have watched a trial for 30 seconds before the word was triggered. I can’t find a good explanation for these instances in the paper, and going by comments like these in the preprocessing script of the study:

# Note: For some reason this didn't work in mutate. No idea why. 
trial_data$time_audio2 = trial_data$Timestamp - as.numeric(audio2_onsets[audio2_trial_number])

it does not seem like the authors fixes these either. It appears they just looked at the timespan after the point of disambiguation.

What do we do here generally? Dig into it and exclude participants that seem weird (& how would that be decided), or just take the data as is?

adriansteffan commented 8 months ago

As suspected, there were very very few timestamps falling below -4000 t_norm, meaning they are most likely artefacts created by experiment mishaps. I have filtered them out and documented this decision in the script.