v01 - Final, edited datasets.
Documented here are final, cleaned datasets prepared by the World Bank based on raw datasets provided by the study researchers.
The description of the difference between raw and edited datasets is taken from "Data Cleaning Guide for PETS/QSDS Surveys" (p.10):
"Each country set includes two data files. The first file, the "raw" data file, presents the data as collected and entered by the survey teams. While field teams do conduct very high-level coherence tests with regards to responses collected, the data contained therein has generally not been thoroughly checked for internal coherence across questions, variable outliers and other such involved data cleaning procedures.
The second file, the "final" data file, has been reviewed in order to ensure consistency both within and across single observations. While the sanctity of data is paramount, such that no changes are made if it cannot be asserted that the edited value is closer to the "true" value than the previous entry, data edits are introduced into the final data set. The list of edits applied are listed in the available Stata 10 © do-file associated with each data set. Furthermore, each do-file includes other tests that were applied to the data set. In addition, basic statistical analysis is applied to variables in order to identify potential statistical outliers. Outlier values that cannot be explained are replaced by missing values in the "final" data set; these changes are reported both in the do-file and in the Data Quality Report.
Finally, independently of the values presented in the questionnaires, missing values are replaced across all "final" data sets to ensure consistency across countries. Following industry best practices, negative 3-digit integers are used in order to ensure there is no confusion between missing values and valid data points. "
"Data Cleaning Guide for PETS/QSDS Surveys" is available in external resources.
