MI (nonparametric / loess)
Mass imputation with a nonparametric (loess) outcome model. Only a small number of covariates is supported by nonprobsvy; here we fit y ~ x1 + x2.
nonprobsvy version: 0.3.0 | R: 4.6.0 | run: 2026-05-24 08:05:57 | commit: 4b2ba9a
| script | var_method | rep_type | num_boot | alpha | n_reps | bias | rmse | mc_se | mean_se | coverage | ci_width |
|---|---|---|---|---|---|---|---|---|---|---|---|
| mi_npar | analytic | NA | NA | 0.05 | 500 | 0.014 | 0.090 | 0.004 | 0.051 | 0.742 | 0.202 |
| mi_npar_boot | bootstrap | subbootstrap | 50 | 0.05 | 100 | 0.015 | 0.084 | 0.008 | 0.079 | 0.900 | 0.308 |
Notes
- DGP:
default. - Analytical SE for loess-based MI uses an asymptotic linearisation that, like the NN/PMM variants, doesn’t fully account for smoothing-window variability. Bootstrap is the safer choice for honest inference.