The psychometric test I have presents 20 test trials to each subject. It does not record the response time (RT) taken toe complete each trial and all I have are the total times taken to complete ALL trials. I am interested in the mean RT take to respond correctly to a trial. What should I do to compute an estimate of this measure from another estimate of th mean RT taken to respond incorrectly to a trial?
To answer it will help knowing why you want the mean given that RT distributions are usually very skewed (e.g., depending on the task, log-normal often fits okay). BUT, to answer your question, the mean would be found by dividing by 20, but that is obvious, so I assume that you are asking for something more than this.
Hello Jimmy,
Hindsight is so perfect! All that needed to happen was to record the RT to each trial...
What could you do, other than collect more data? If you are willing to assume that:
1. Response time to correctly performed tasks is consistent regardless of task;
2. Response time to incorrectly performed tasks is consistent regardless of task;
3. This relationship (of #1 to #2) is consistent across persons (more or less);
Then, record Number right, Number wrong, and total time to completion as three variables for each case. Regress time to completion on NR and NW, and determine what estimated coefficients for each imply about whether encountering a task one can perform yields a different RT than does encountering a task one can not perform. Do note that this approach isn't ideal, since there are most likely differences in mean RT across persons, regardless of condition, and that is simply adding noise to the data set.
Good luck with your work.
- Badges
- Science topic
- Similar topics
- Psychology
- Psychometrics