Matching in Surveys
Ehsan Karim 
Video content
Slides
Summary of topics discussed
Tabular summary
| Suggestions | PS Model | Outcome Model | Source |
|---|---|---|---|
| PS stratification | No weights | Use weights (no comment on strata / cluster) |
Zanutto, E. L. (2006) |
| PS as covariate | Use weights + strata + cluster as predictors | Use weights + strata + cluster as design features | DuGoff, E. H., Schuler, M., & Stuart, E. A. (2014) |
| Weighting | Use weights (+ strata + cluster) as design features | Use weights (+ strata + cluster) as design features | Ridgeway, G., Kovalchik, S. A., Griffin, B. A., & Kabeto, M. U. (2015) |
| Matching | Use original weights + strata + cluster as design features | Use weights + strata + cluster as design features | Austin, P. C., Jembere, N., & Chiu, M. (2018) |
| Matching | Balance should be deciding factor | Use weights + strata + cluster as design features | Lenis, D., Nguyen, T. Q., Dong, N., & Stuart, E. A. (2017) |
Reasonable approach (my suggestion; similar to Lenis et al. 2017):
- Use if it provides best balance (e.g., measured by SMD) for PS model building. Often applying design features in the PS estimation is not possible due to the shortcoming of the software available for the machine learning approach chosen to estimate PS.
- Must use all design features in the outcome modelling for population estimates |
References (Optional)
- Zanutto, E. L. (2006). A comparison of propensity score and linear regression analysis of complex survey data. Journal of data Science, 4(1), 67-91.
- DuGoff, E. H., Schuler, M., & Stuart, E. A. (2014). Generalizing observational study results: applying propensity score methods to complex surveys. Health services research, 49(1), 284-303.
- Lenis, D., Nguyen, T. Q., Dong, N., & Stuart, E. A. (2017). It’s all about balance: propensity score matching in the context of complex survey data. Biostatistics. 2019 Jan; 20(1): 147–163.
- Austin, P. C., Jembere, N., & Chiu, M. (2018). Propensity score matching and complex surveys. Statistical methods in medical research, 27(4), 1240-1257.