|
1:00 pm Registration
1:30 pm Invited Speaker: Brad Smith
Department of Psychology
University of South Carolina
'Evaluating the Precision and Efficiency of Binge
Drinking Cutoffs for
Predicting Alcohol-Related Problems Using the Methods of Meehl & Rosen
(1955)'
Self-reported data from 1,479 college students were used to compare
cutoffs for predicting alcohol-related problems. We compared "binge
drinking" (BD) cutoffs (i.e., five or more drinks for males and four or
more drinks for females on one occasion) to cutoffs that achieved the
optimum balance of specificity and sensitivity on Receiver Operating
Characteristic (ROC) Curves. Only 3 of 14 alcohol-related problems had
cutting scores between 1 and 15 drinks that met the efficiency criteria
of Meehl and Rosen (1955). With the exception of a high frequency
problems scale, ROC cutoffs explained more variance in alcohol-related
problems than BD cutoffs. These findings highlight the importance of
considering base rates, efficiency, and precision when using cutting
scores.
2:30pm Student Paper Competition:
(Select the title of the talk to view the abstract)
4:15 pm Invited Speaker: David Smith
Office of Biostatistics and Bioinformatics
Medical College of Georgia
'Local and Bayesian Optimal Designs in Binary Bioassay'
When designing a bioassay the number of doses and numbers of (say)
insects assigned to those doses profoundly effects the likely precision
(variance) of the estimates of parameters and other criteria of design
performance. This talk will be a review of current work and work already
done looking at performance in binary bioassays from an optimal design
perspective. Issues such as how many doses; where they should be placed;
and the numbers (weights) at those doses will be discussed.
5:15 pm Election of Officers
| 