Page | 33 The group matching will be accomplished using propensity score matching with logistic regression. For each participant, a single, composite propensity score will be created to reflect the status of the individual with respect to initial language ability, gender, race, ethnicity, and economic level. For each intervention group student, we will identify a comparable student in the comparison group with a propensity score within 10% of the intervention group student. Student matches will be sampled without replacement. The number of matched pairs cannot exceed the total size of the larger of the two study groups. Depending on the characteristics and distributions of the two study groups, we would expect a final sample size of about 80% of this maximum size. So, if we were to successfully recruit 1,000 comparison group students and 500 intervention group students, we would estimate a final sample before attrition of approximately 400 treatment and 400 comparison group students (i.e., a total sample size of 800). Comparability of Treatment and Comparison Groups We will compare the characteristics of the matched intervention group and comparison group obtained through propensity score matching. We will examine the mean and standard deviation of the initial language skills for the intervention and comparison group samples targeting a diPerence in mean scores equal to no more than one half standard deviation and ideally one quarter of a standard deviation. The number and percentage of students in each group with respect to gender, race, ethnicity and economic status will be compared. Attrition The number of students in the final analysis sample may be reduced from the original matched sample for a number of reasons. For example, a student may relocate out of the school during the course of the study, or the student may not take the posttest due to absence or other factors. We anticipate attrition of between 10% and 20%, resulting in a final estimated analysis sample of approximately 640 to 720 students. SEG will provide final attrition counts both within and across study groups. Final Analysis Sample After accounting for attrition, approximately 700 students will likely remain in the final analysis sample, 350 students in the treatment group, and 350 students in the comparison group.
RkJQdWJsaXNoZXIy MjUyNzA0NQ==