4 Study Population and Sampling The sample of students participating in this study was drawn from two sources: the population of grade one, two, and three students in the school district studied in this research and a national sample of approximately 3,200 grade one, two, and three MLs. We drew the district sample from a file containing all students at the three grade levels, including approximately 278 MLs receiving instruction incorporating the treatment and the remaining 1,173 non-MLs who did not receive instruction incorporating Connect. The treatment group was composed of the 278 MLs receiving the instruction incorporating Connect; the comparison group was composed of a sample of the group of remaining non-MLs at the same grade levels not receiving instruction incorporating Connect. The national sample was drawn from the sample of MLs completing the MAP assessment in districts across the US. Study Sample for Question 1 The study sample used to answer Question 1 was the total treatment group of 278 MLs receiving instruction incorporating Connect. Four students were lost to attrition. These four students took the pretest but left the district or otherwise did not have a posttest score available. Study Sample for Question 2 The 278 ML student sample comprising the treatment group for Question 1 was augmented to include a national sample of MLs. NWEA provides a large research database containing nearly 60,000 students nationwide. The approximately 3,200 students classified as ML throughout their grade one, two, and three experience drawn from the NWEA research database were used as a basis for answering Question 2. Study Sample for Question 3 The sample of 310 students used to answer Question 3 included 155 students drawn from the treatment group and 155 students drawn from the comparison group, a sample of the remaining same grade level students not receiving instruction incorporating Connect. Creating Matched Samples Given that random sampling was not possible for this study, SEG Measurement used a matched groups quasi-experimental design. For each student in the treatment group, an individual from the control group similar in ability and background was selected. We created matched treatment and control groups using propensity score matching, an advanced statistical technique that permits matching based on multiple factors. Using a logistic regression-based propensity score model, students were matched on a composite propensity score created from initial reading ability (MAP English language assessment pretest scores), their sex, ethnicity, and free or reduced-cost lunch eligibility. Sampling was done without replacement and the composite propensity score for the students in the matched pair was required to be within 10% of each other.
RkJQdWJsaXNoZXIy MjUyNzA0NQ==