I have been reading Journal articles that have standardized RLs using log transformations or other transformations.
Two questions:
a) Is standardization needed/recommended?
b) If so, any suggestions on what method to use to standardize?
Thanks
Printable View
I have been reading Journal articles that have standardized RLs using log transformations or other transformations.
Two questions:
a) Is standardization needed/recommended?
b) If so, any suggestions on what method to use to standardize?
Thanks
We are not really in a position to declare any kind of standard transformation. It depends on a lot of factors involving the patterns in your data (uniformity and shapes of distributions and such), personal taste, tradition in your area, and what kind of effects you are looking for. Best recommendation I would make is to read the [I]Results[/I] sections of RT studies that have achieved classic status and articles that focus on RT methodology. Here is a good start:
Fazio, R. H. (1990). A practical guide to the use of response latency in social psychological research. In C. Hendrick & M. S. Clark (Eds.), [I]Review of Personality and Social Psychology: Vol. 11[/I]. [I]Research Methods in Personality [/I][I]and Social Psychology[/I] (pp. 74-97). Newbury Park, CA: Sage Publications.
But with that said, off the record, at the very least it's good to explore your data using a variety of transformations as doing so can help you see better what's actually going on and why your regression equations are turning out the way they are. I, personally, would look at my raw data and then look at it with outliers capped at 2.5 SD from the mean. Overall mean? Mean from their particular cells? Yup, there are many ways to go. And why 2.5 SDs? Yup--relatively arbitrary, no? And how about those log transformations--yes, a great thing to try. I would typically do this after controlling for outliers but you could argue the other way around. Oh, and be sure to keep track of how many trials you are capping or designating as "bad trials". The more you have to change individual data points or apply transformations differentially to different cells, you are going to be looking at red flags for reviewers (as opposed to global transformations like a nice single log transform that is applied to all scores). Anyways, you can get a million answers on stuff like this. Personally, I would check out articles by Fazio and Bargh and the classics from the cognitive domain. But then again, remember this is all just one guy's off the top of his head thoughts.