2.2.3. Scoring and Data Treatment
The data treatment followed the procedure of Cepulic et al. (2018) and
was conducted via a Python script, using Numpy and Pandas packages. For
obtaining accuracy data, we excluded fast guesses with reaction times
(RT) < 200 ms and calculated the average accuracy for each
block. For calculating RT data, we excluded incorrect trials (15.04% of
trials in the whole dataset). Then, the data was winsorized (i.e., Tukey
correction; Tukey, 1977) at the within-person level, that is, for each
block, all RTs longer than the third quartile plus 1.5 times the
interquartile range were set to this limit value (7.9% of trials in the
whole dataset). Based on the winsorized data, average RTs for each
condition in each participant were calculated. We also inspected the
averaged data across all conditions for interindividual outliers based
on the same criterion; no participants fell outside of this limit.
Finally, the mean RTs per condition and participant were transformed by
multiplying the reciprocal of the average RT (ms) by 1000 (1000/RT) to
normalize the distribution. The transformed RT data reflect the number
of correct responses per second, that is, larger values indicate greater
speed.