MI (GLM)
Mass imputation with a GLM outcome model. The non-probability sample fits y ~ x1 + x2 and the predictions are aggregated over the probability sample.
nonprobsvy version: 0.3.0 | R: 4.6.0 | run: 2026-05-24 08:18:26 | commit: 4b2ba9a
| var_method | num_boot | alpha | n_reps | mean_se | coverage | ci_width |
|---|---|---|---|---|---|---|
| analytic | NA | 0.05 | 500 | 0.077 | 0.944 | 0.302 |
| analytic | NA | 0.10 | 500 | 0.077 | 0.898 | 0.254 |
| bootstrap | 200 | 0.05 | 500 | 0.070 | 0.922 | 0.273 |
| bootstrap | 200 | 0.10 | 500 | 0.070 | 0.854 | 0.229 |
Notes
- DGP:
default(linear outcome,y = 1 + x1 + 0.5 x2 - 0.3 x3 + N(0,1)). - The outcome model is correctly specified for this DGP.
- Analytical SE for GLM-based MI uses the standard linearisation argument and should yield well-calibrated CIs at the nominal level.