X hits on this document

188 views

0 shares

51 / 59

# Weighting

The general purpose of weighting is to scale up the sample estimates to represent the target survey population. For SASS, a base weight (e.g., the inverse of the sampled teacher’s probability of selection) is used as the starting point. Next, a series of nonresponse adjustment factors are calculated and applied using information from the 2003-04 SASS nonresponse bias analysis and information about the respondents known from the sampling frame data. Finally, for some files, a ratio adjustment factor is calculated and applied to the sample to adjust the sample totals to the frame totals. The product of these factors is the final weight for each SASS respondent, which appears as DFNLWT on the SASS Public School District data file, AFNLWGT on all SASS Principal data files, SFNLWGT on all SASS School data files, TFNLWGT on all SASS Teacher data files, and MFNLWGT on all SASS Library Media Center data files.

The counts in table 1 do not necessarily match the frame counts because some cases in the frame were found to be ineligible (i.e., out-of-scope) and because not all data files (e.g., principal or library media center) are post-stratified to match the frame counts.

Variance Estimation

In surveys with complex sample designs, such as SASS, direct estimates of sampling errors that assume a simple random sample typically underestimate the variability in the estimates. The SASS sample design and estimation include procedures that deviate from the assumption of simple random sampling, such as stratifying the school sample, oversampling new teachers, and sampling with differential probabilities.

One method of calculating sampling errors of complex sample designs is replication. Replication methods involve constructing a number of subsamples (i.e., replicates) from the full sample and computing the statistic of interest for each replicate. The mean square error of the replicate estimates around the full sample estimate provides an estimate of the variance of the statistic. Each SASS data file includes a set of 88 replicate weights designed to produce variance estimates. The set of replicate weights for each file should be applied to the respondents in that file. The replicate weights for SASS respondents are DREPWT1-DREPWT88 for districts, AREPWT1-AREPWT88 for principals, SREPWT1-SREPWT88 for schools, TREPWT1-TREPWT88 for teachers, and MREPWT1-MREPWT88 for library media centers.

B-11

 Document views 188 Page views 188 Page last viewed Thu Jan 19 04:42:58 UTC 2017 Pages 59 Paragraphs 2043 Words 17045