Date created: 2017-03-23 09:39 PM | Last Updated: 2018-02-23 08:25 PM

Category: Project

Description: Researchers wishing to use our approach to assess the robustness of their own findings to the presence of potentially mischievous responders can either (1) recreate (or create some modified version of) our procedure and the resulting weights or (2) download the weights we created from our main analysis from this site. The data are available in csv format for our final analytic samples (i.e., anyone receiving a mischievousness score) and can be merged with the 2015 CDC data using the record ID. The file contains 26 weights for each of the 148,960 observations in the State and District study and the 14,612 observations in the National study. One weight (labeled simply “weight”) is the original sampling weight provided by CDC. The other 25 weights are labeled “dropboost_XX,” where the XX corresponds to the proportion of the unweighted sample remaining after dropping the (100-XX)% of mischievous responders. For example, dropboost_99 should be used as the sampling weight if wishing to remove only the top 1% of individuals on the mischievousness scale, whereas dropboost_80 should be used to remove the top 20% on the mischievousness scale. The CDC periodically updates the YRBS datasets it makes available and renumbers the IDs in the process. Note that the weights uploaded to the Open Science Framework website were based on analyses run on the 2015 YRBS dataset that was current on the CDC website as of August 4, 2017.