The GSS RBS sample included 2,500 departments selected from a list of GSS departmental contacts. An important challenge in selecting a sample for the RBS was that the study had three units of analysis: respondent, department, and institution.
Stratification by reporting type
The first level of stratification was the reporting type, the number of departments and types of populations represented by a GSS respondent. The reporting type appeared to have the greatest potential impact on both response behavior and postdoc data quality. The strata included:
- One department with postdocs only:
Respondents who reported for one department that only has postdocs
- One department with graduate students only:
Respondents who reported for one department that only has graduate students
- One department with both:
Respondents who reported for one department that has both postdocs and graduate students
- Multiple Departments:
Respondents who reported for more than one department with:
- Postdocs only
- Graduate Students only
- Both postdocs and graduate students
Sampling based on survey burden
RBS used a Probability Proportionate-to-Size (PPS) sampling strategy. The Measure of Size (MOS) was the total number of postdocs and/or graduate students in the department for which the GSS respondent reported. GSS respondents with a high MOS had a higher probability of being selected to the RBS sample.
For the respondents who were responsible for reporting for multiple departments, we selected one department on which the respondent reported. The sample was not balanced by reporting type when it was selected. To ensure that the sample was representative of the GSS population, post-stratification weights were applied.
Sample allocation
The first sample and the replicate were combined when analyzing the data. An indicator variable was generated to distinguish respondents selected to either group.
|