+x+x+x+xI would like to add a few general remarks, if I may:
For the vast majority of RT tasks used in psychological research, there is no canonical, universally agreed upon, cookbook data analysis procedure. The fact that such a procedure exists for the IAT is an absolute exception.
RT data analysis involves lots of (sometimes rather subjective) decisions. Here's a completely arbitrary list of some recurring issues:
- Traditionally, people compute aggregate measures and then run ANOVAs or regression models on those aggregates.
- RT distributions typically aren't symmetrical, hence some people prefer using medians instead of means.
- Again others apply transformations to RTs prior to analysis to make the distributions more symmetrical (typically log or reciprocal). Others frown on such transformations.
- There are a bazillion ways to screen for and eliminate outliers in your data. Some people use SDs (commonly +/- 2 or 3), others prefer Winsorizing, etc.
- Proponents of more modern methods don't aggregate data at all. Instead they perform hierarchical / multilevel modelling on the raw data.
- There are even more specialized methods for explicitly modelling RT distributions (ex-Gauss, Weibull), or procedures like Ratcliff's diffusion model (to analyze accuracy and RT concurrently).
- In case of Stroop tasks, some people prefer to analyze difference scores (as indicated in Sean's response), others would prefer to simply include (in)congruence as a within-subjects factor in an ANOVA.
I could go on forever. All that is to say: If you don't know how to analyze data for a given task, I'd recommend you
- Acquire a general background in RT data analysis and all the steps involved in it. There's an ample body of literature on every single one.
- Refer to the relevant, task-specific literature to identify the common analysis choices and examine the reasoning behind them.
Picking a suitable data analysis plan requires lots of (hopefully) informed choices, and nobody can make these for you.
Hi there,
This is a
very basic question, but what is the difference between "expressions.meanrt" and "expressions.interference" - I'm having a look at my summary file, and they are both the same number (i.e., 524.55 for control, 627 for food words).
What is the difference between the two? does one of them take into account mistakes made?
Thanks in advance for your help!
stefania
At a minimum, you need to tell me which exact Stroop script you are referring to (there are many), and if the script does not come from the millisecond.com library or has been extensively modified by somebody else, I cannot answer the question.
For scripts from the library, you'll generally find data output explained in the comments inside the script.
The task is the Alcohol stroop task, and I have modified it (not extensively though) it.
Verbal or keyboard input? The Inquisit 5 or the Inquisit 4 version of the script? What did you modify, in brief?
I'm asking because, for example, in the keyboard version for Inquisit 5 ->
https://www.millisecond.com/download/library/v5/stroop/alcoholstroop/alcoholstroopwithkeyboardinput.iqzip there is no such thing as "expressions.interference" -- there are three expressions, namely
expressions.interference_alcohol, expressions.interference_negativewords, expressions.interference_positivewords
and they're calculated like so:
<expressions>
/meanRT_Alcoholwords = values.sumrt_alcoholwords/values.
correct_alcoholwords/meanRT_negativewords = values.sumrt_negativewords/values.
correct_negativewords.../Interference_alcohol = expressions.meanrt_alcoholwords - expressions.meanrt_neutralwords
/Interference_negativewords = expressions.meanrt_negativewords - expressions.meanrt_neutralwords
/Interference_positivewords = expressions.meanrt_positivewords - expressions.meanrt_neutralwords
</expressions>
From that you can also see that the mean RTs are based on correct responses only, as detailed in the comments:
/meanRT_Alcoholwords: mean latency of correct responses to Alcohol words across blocks
/meanRT_negativewords: mean latency of correct responses to negative words across blocks
/meanRT_neutralwords: mean latency of correct responses to neutral words across blocks
/meanRT_positivewords: mean latency of correct responses to positive words across blocks
Interference Scores:
"interference scores were calculated by subtracting each participant’ s mean reaction time to
the neutral stimuli from that participant’ s mean reaction time to alcohol, positive and negative
stimuli, respectively."
/Interference_alcohol: Difference btw. mean reaction time to alcohol and neutral words (latencies of correct responses only)
/Interference_negativewords: Difference btw. mean reaction time to negative words and neutral words (latencies of correct responses only)
/Interference_positivewords: Difference btw. mean reaction time to positive words and neutral words (latencies of correct responses only)