Can researchers produce state-level estimates with the ECLS-K data?
No, the ECLS-K sample was designed to support national and regional estimates. It was not designed to estimate characteristics of children, teachers, or schools at the state, county, or city level.
Can you use the ECLS-K data to produce estimates that are nationally representative of school and teacher characteristics?
Yes, but only the ECLS-K kindergarten data will support such estimates. The base-year (i.e., kindergarten) school sample is nationally representative of schools that educate kindergartners. A separate school-level data file with school weights is included on the longitudinal kindergarten through eighth grade data file. During the kindergarten year, the ECLS-K sampled all kindergarten teachers in each of the ECLS-K schools. Data from this nationally representative sample of kindergarten teachers also are available as a separate file with teacher weights on the longitudinal kindergarten through eighth grade file. The ECLS-K data do not, however, support teacher-level or school-level estimates in first, third, fifth, or eighth grades. After kindergarten, teachers and schools were only included in the study if they educated one or more ECLS-K children. Therefore, no teacher-level or school-level weights are provided after the base year.
Do I need to use weights for my analyses?
Weights are used to adjust for disproportionate sampling, survey nonresponse, and undercoverage of the target population when analyzing complex survey data. They also are necessary to produce national-level estimates of the ECLS-K cohort, of kindergarten teachers in 1998-99, and of schools educating kindergartners in 1998-99.
Do I need to use the same weight for all analyses?
Researchers are encouraged to use the same weight throughout all analyses in a publication or paper, even when there is a different ideal weight for each analysis. Weights are assigned to cases with valid data associated with the component(s) contributing to the weight. Selecting different weights within the same publication or paper results in each analysis being run with a different analytic sample (i.e., the exact cases contributing to the analyses).