报告一：

Title: Integrating ideas from statistics and engineering into behavioral intervention science

Speaker: Linda M. Collins (ThePennsylvaniaStateUniversity)

Time: 10:00-10:50am, June 29^th.

Place: Rm216, GSM New building

Abstract: Behavioral intervention programs are being used increasingly for prevention and treatment of disease and promotion of health and academic achievement. Examples include smoking cessation intervention programs; programs that help people to maintain a healthy weight by establishing better eating and exercise habits; interventions for treatment of chronic diseases such as diabetes; and school-based programs to improve reading comprehension. Historically behavioral interventions have been developed a priori and then evaluated against a control group, usually by means of a two-group randomized experiment. This approach has enabled investigators to determine whether an intervention has a statistically significant effect and to estimate the effect size. However, it has only limited utility in scientific development and improvement of behavioral interventions. This presentation will discuss research aimed at integrating ideas from statistics and engineering into behavioral intervention science. The objective is to enable behavioral scientists to optimize behavioral interventions, that is, to engineer interventions to meet specific criteria of effectiveness and efficiency.

茶歇：206室外，10：50-11：00am

报告二：

Title: Clinic and ambulatory BP measurement: A valid and cost-effective approach using the two-method measurement design

Speaker: John W. Graham (ThePennsylvaniaStateUniversity)

Time: 11:00-12:00am, June 29^th

Place: Rm216, GSM New building

Abstract:

Researchers who study blood pressure (BP) have two measurement options: collect BP measures (1) using standard clinic-style methods (ClinBP), or (2) using Ambulatory BP (AmbBP) procedures. ClinBP is relatively inexpensive, but is known to be affected substantially by false positive, and false negative indications of hypertension. AmbBP is highly valid but costs 5 times as much to implement. Two-method measurement, a planned missing data design (Graham et al., 2006), offers a nice solution to the dilemma faced by BP researchers. With this approach, all subjects are measured using ClinBP, but only a small random sample of subjects also receive the more expensive, but more valid AmbBP measures. Results of our simulations showed that the two-method approach was as valid as, but much more powerful than using the AmbBP approach. For the same overall study costs, the two-method measurement approach yielded an effective N that was nearly double the N for the complete cases (AmbBP+ClinBP) design costing the same.