Interesting problems often arise in data analysis. These often drive me to develop new techniques for statistical or mathematical modeling. In a major program of such research, funded by a grant from the National Science Foundation, I am developing efficient strategies for assessing difficult-to-test subjects. The application that led me to this line of research was testing the hearing of some of the less cooperative subjects in the world: infants. Because infants can easily fatigue, lose interest, or become fussy, it is critical to choose one’s stimuli very carefully, and to extract as much information as possible from the infant’s pattern of results. In solving this problem, I developed a general procedure rooted in information theory for choosing stimuli likely to be maximally informative in classifying subjects (in this case, as normally hearing or as having a particular hearing loss). These methods are very broadly applicable, and I am now extending this work to the development of efficient measurement techniques in other realms (e.g., psychophysics, assessment of children’s vocabulary).

My next application of the adaptive-methods work (the evaluation of vocabulary in young children) has just been funded by the National Institutes of Health.

Another example of statistical/mathematical research is work by myself and Craig Mason on adapting statistical techniques common in epidemiological research so that they are more appropriate for many areas of psychological research.

And, yes, that is a picture of my calculator (one of them, at least).

Hey, buddy, wanna buy a stellated icosahedron?