Adaptive design of glove sensitivity
Test how gloves affect fingertip sensitivity. 5 types of gloves. Fingertip sensitivity is measured by a dome with 9 different distances between grooves.
Participants were always started on a middle distance dome and the goal is to determine which orientation the dome is in. The participant gets 10 attempts for each distance. Then if the participant does well, the distance is decreased making the test more difficult or, if the participant did well, the distance is increased making the test easier. Typically stopping after either the participant gets 8 or above or 3-4 different domes.
- Just use the data for the start groove distance assuming everybody started with the same groove distance.
- Summarize each participant-glove by the end groove level and us this as the response. The issue here is that the data were not collected consistently.
- Take a weighted average of the dome groove distance with a weight that is related to the proportion correct.
- Fit a mixed effect logistic regression model with fixed effects for glove and distance (continuous) and fixed/random effects for student. This model assumes there is no order effect for distance and gloves and the missing data are missing at random.
For next time:
- Randomly assign the order of gloves within each subject
- Have a procedure for distance and stick to it.
blog comments powered by Disqus