What would be considered as reaction time outliers in the Serial Reaction Time Task (SRTT)?
Some previous studies suggested RT below 300 ms and above 4000 ms should be excluded.
1) Is there a more flexible range for outliers specially for fast RTs? for example 200 ms or less?
2) is it possible to consider fast RTs (e.g., 300-150 ms) in the analysis? Because these fast RTs are indicative of learning and excluding them just affect the results.
Dear @ M.A. Salehinejad
I would not recommend having a fixed threshold for exclusions for the whole experiment. I think the most sensible strategy is to have a different threshold for each participant and for each learning block equivalent to +/- 3 Median Absolute Deviations. As a rule of thumb this procedure should not exclude more than 5% of the data. If you want, maybe my student Johanna Sánchez-Mora can provide you with some sample scripts for doing that.
Another possibility, if you are interested in the central tendency of the RT distribution, is to not remove any outlier and compute the harmonic mean which is robust to long RTs (see Ratcliff R. Methods for dealing with reaction time outliers. Psychol Bull. 1993 114(3):510-32.).
- Badges
- Science topic
- Similar topics
- Philosophy
- Philosophy Of Science
- Reasoning