Figure 2 : Matching accuracy according to ε+.

From: Probabilistic record linkage of de-identified research datasets with discrepancies using diagnosis codes

Figure 2

The figure shows the performance of the matching according to various value of the hyper-parameter ε+ which models the probability of discrepancy between the two datasets. The top panel displays results when true value of ε+ is 3%, while the bottom panel is for a true value of 20%.