Page | 32 Assessment We are assuming that the participating district(s) will use a commercially available, thirdparty language skills assessment such as described above. These assessments generally have high reported reliability and validity. SEG will obtain the reliability and validity data to the extent available and include this in the final study documentation. The assessment overall and the subtest data will be examined for the study. Study Population and Sampling VHL is currently engaged in recruiting and evaluating school districts to participate in the study. We are seeking between one and two districts and three or more schools for the study. We are seeking approximately 1,500 students in two grades within the grades 9−12. This reflects approximately 50−75 classes/teachers. The target sample size of 1,500 is designed to have suPicient statistical power to detect a minimum ePect size of .15 between the two study groups if present, with a minimum level of significance requirement of .05 and power at or above .80. Power for subgroup comparisons is lower due to smaller sample sizes. More detail on the steps in the sampling process are provided below. As noted earlier, the study may employ an experimental design (associated with ESSA Level 1 evidence). This would include random sampling of students with each study participant randomly assigned to either the intervention group or the comparison group. This would maximize the likelihood that the two study groups are similar in composition particularly with respect to initial language skills, gender, race, ethnicity, and economic level. There are many challenges to random assignment in this setting, so employment of a quasiexperimental design (associated with Level 2 evidence) may be considered. Here, an intervention and comparison group would be established, but study participants would not be randomly assigned to the two groups. While we plan to take reasonable ePorts to select groups that are comparable in order to achieve comparability, we would likely match the two groups. Creating Matched Samples If random sampling is not possible for this study, SEG will use a quasi-experimental design with matched groups. Group matching will be accomplished using propensity score matching. For each student in the intervention group, we will identify and select an individual from the comparison group who is similar in initial language ability, gender, race, ethnicity, and economic level.
RkJQdWJsaXNoZXIy MjUyNzA0NQ==