|
Intraclass Correlation Coefficients and Effect Sizes for Planning School-Randomized Experiments
|
| Presenter(s):
|
| Paul R Brandon, University of Hawaii, Manoa, brandon@hawaii.edu
|
| George Harrison, University of Hawaii, Manoa, georgeha@hawaii.edu
|
| Brian Lawton, Independent Consultant, blawton@hawaii.edu
|
| Jerald Plett, Hawaii Department of Education, jerald_plett/sas/hidoe@notes.k12.hi.us
|
| Abstract:
The power analyses conducted for group-randomized experiments require estimating minimum detectable effect sizes (MDES) and intraclass correlation coefficients (ICC), among other statistics. The estimates should be based on previous empirical studies. In the first part of our study, we present ICCs and MDESs for the Hawaii State Assessment reading test and mathematics test for seven grades over nine years. The results provide an empirical basis for planning school-randomized experiments in Hawai'i and elsewhere. In the second part of the study, we present distributions of Hawai'i public-school reading and mathematics between-school effect sizes over the seven grades and nine years. The distributions help show the likelihood that researchers can achieve the MDESs that we calculate in the first part of the study. We provide the SAS macros developed to conduct the analyses. Our methods can be adopted by any educational jurisdiction, thereby helping guide evaluation studies nationwide.
|
|
Longitudinal Multivariate Analysis of Ecological Theory to Increase Highway Safey and Reduce Fatalities and Serious Injuries
|
| Presenter(s):
|
| Robert Seufert, Miami University, seuferrl@muohio.edu
|
| Abstract:
To obtain federal highway funds, states must annually collect observational data on seat belt use. Also, states annually compile crash report data indicating that fatalities and serious injuries are directly or indirectly influenced by factors, including vehicle miles traveled, roadway type, seat belt use, speed, alcohol impairment, vehicle type, motorcycle and helmet use, cell phone use, and other distractions. This research analyzes data from both databases for 2006 through 2010 and tests an ecological model on the interrelationship between variables within multiple environments. The theory depicts 'a nested structure of environments' with a complex web of causation and context for implementing effective interventions to prevent highway fatalities and serious injuries. The theory is tested with statewide data samples, each containing approximately 20,000 vehicle occupants, and complete annual crash report data. In summary, the ecological theory and multivariate analysis clarifies opportunities for effective interventions to prevent highway fatalities and serious injuries.
|
| |