nonprobsvy: An R Package for Modern Methods for Non-Probability Surveys1 months ago
Introduction | Software for non-probability samples | Methods for non-probability samples | Basic setup | Prediction-based approach | Prediction estimators | Mass imputation estimators | Variance estimators for the prediction approach | Inverse probability weighting | Variance estimators for the inverse probability weighting approach | Doubly robust approach | Parameters estimated separately | Minimization of the bias for doubly robust methods | The goal of this approach is to minimize ${a!\operatorname{bias}(\boldsymbol\gamma,\boldsymbol\beta)}^2$, which leads to the following system of empirical estimating equations:$$\left(\begin{array}{c}\sum_{i=1}^NI_{\text{NP}, i}\frac{\dot{\boldsymbol\pi}\left(\boldsymbol{x}_i,\boldsymbol{\gamma}\right)}{\pi\left(\boldsymbol{x}i,\boldsymbol{\gamma}\right)^2}\left{y_i-m\left(\boldsymbol{x}i,\boldsymbol{\beta}\right)\right}\[1.2em]\sum{i=1}^N\frac{I{\text{NP}, i}}{\pi\left(\boldsymbol{x}_i,\boldsymbol{\gamma}\right)}\dot{m}\left(\boldsymbol{x}_i,\boldsymbol{\beta}\right) | \sum_{i \in S_{\text{P}}}d_{\text{P}, i}\dot{m}\left(\boldsymbol{x}_i,\boldsymbol{\beta}\right)\end{array}\right) | Variance estimators for the doubly robust approach | Variable selection algorithms | The main function and the package functionalities | The nonprob function | Controlling the type of estimators | Controlling variance estimation | Data analysis example | Description of the data | Estimation | Propensity score approach | The doubly robust approach | Comparison of estimates | Advanced usage | Bootstrap approach for variance estimation | Classes and S3 methods | Summary and future work | Acknowledgements | Appendix: Algorithms for the MI-NN and MI-PMM estimators | References