I'm running a cognitive psychology experiment right now which is producing
some really exciting results, and which the subjects find fun and
challenging. However, I get the occasional subject who stops to comment, or
to giggle, and this inflates response time for that trial. I just had a
guy, in fact, who - despite the instruction to the contrary - stopped from
time to time to ponder like a chess player. It seems to me that rounding
response times to a reasonable outlying value would be one approach. I
could delete such responses as well, but they are generally correct
responses, and this would seem to me to introduce greater bias, and require
an unbalanced ANOVA model. Any comments?
David Klein
|