Working with survey data
My undergraduate dissertation used IBM SPSS Statistics. I now use R for survey preparation and analysis in the current research project.
Working with CGSS involves translating survey responses into explicit sample rules: identifying completed qualifications, distinguishing current from previous occupations, coding adolescent origin and handling missing parental education. For the income comparison, I distinguish zero income from unavailable amounts and keep the same respondents in both outcome models.
Occupational measurement
I convert ISCO occupational codes into ISEI status and SIOPS prestige scores. I check the level of detail in the recorded occupation codes and compare scoring methods on the same respondents. This allows me to assess whether an association depends on how occupations are valued.
For teaching, I compare models that pool adult and regular qualifications with models that distinguish them, keeping the respondents the same.
Modelling and sensitivity
I use ordinary least squares for occupational status and log labour income, and linear probability models for teaching. The analysis considers interactions, alternative samples, survey weights and province indicators, alongside HC3 and province-clustered standard errors. These comparisons help distinguish patterns that persist from estimates sensitive to measurement or specification.
A documented R workflow
Package versions are recorded with renv. Checks using synthetic data test coding rules and selected statistical calculations. The workflow also checks saved-table consistency and file hashes.
The saved results can be checked for internal consistency without the survey file; reproducing the estimates requires access to the microdata.