Fig. 5
From: Survival analysis under imperfect record linkage using historic census data
Empirical bias for \({\widehat{M}}_{00}\), \({\widehat{M}}_{01}\), \({\widehat{M}}_{10}\) and \({\widehat{M}}_{11}\): (1) MCAR, (2) MAR with both \({T}_{i}\) and \({R}_{i}\) dependent on \({X}_{i}\) only, (3) MAR with \({T}_{i}\) dependent on \({X}_{i}\) and \({Z}_{i}\), while \({R}_{i}\) depends on \({X}_{i}\) only, (4) MAR with both \({T}_{i}\) and \({R}_{i}\) dependent on \({X}_{i}\) and \({Z}_{i}\)