Table 1 Weighting schemes for the proposed missing data methods
From: Survival analysis under imperfect record linkage using historic census data
Weights for unequivocal matches | Weights for equivocal matches | Weights for nonmatches |
|---|---|---|
Complete case (CC) | ||
1 | 0 | 0 |
Inverse probability weighting of all matches (IPW) | ||
\(\frac{1}{{\text{Pr}}\left({R}_{i}=1|{X}_{i},{Z}_{i}\right)}\) | \(\frac{1}{{\text{Pr}}\left({R}_{i}=1|{X}_{i},{Z}_{i}\right)}*{p}_{i,match}\) | 0 |
Censoring nonmatches at \({{\varvec{W}}}_{{\varvec{i}}}\) (CENS) | ||
1 | \({p}_{i,match}\) | 1 (for \({W}_{i})\) |
Multiple imputation for equivocal matches and nonmatches (MIRM and MICS) | ||
1 | \({p}_{i,match}\)(for \({T}_{i}\)) \(1-{p}_{i,match}\) (for \({T}_{i}^{imp}\)) | 1 (for \({T}_{i}^{imp}\)) |