Skip to main content
. Author manuscript; available in PMC: 2024 Sep 12.
Published in final edited form as: Proc Mach Learn Res. 2024 May;238:4195–4203.

Table 2:

Performance comparison of methods on the CPAD dataset under MCAR, MAR, and MNAR mechanisms. We report the mean and standard error for five trials.

Methods MCAR MAR MNAR

MAE RMSE MRE MAE RMSE MRE MAE RMSE MRE

MRNN 0.233(0.001) 0.478(0.013) 33.5%(0.001) 0.233(0.001) 0.672(0.030) 38.5%(0.001) 0.246(0.001) 0.723(0.123) 44.4%(0.003)
RITS 0.259(0.001) 0.547(0.009) 40.3%(0.002) 0.252(0.001) 0.654(0.024) 41.8%(0.001) 0.265(0.002) 0.730(0.122) 47.7%(0.002)
BRITS 0.239(0.001) 0.532(0.008) 37.1%(0.001) 0.233(0.001) 0.629(0.019) 38.7%(0.001) 0.244(0.001) 0.710(0.117) 44.1%(0.003)
CSDI 0.206(0.006) 0.440(0.004) 30.2%(0.009) 0.207(0.002) 0.640(0.025) 34.3%(0.003) 0.219(0.001) 0.679(0.134) 39.5%(0.003)
SADI 0.190(0.001) 0.416(0.005) 28.0%(0.002) 0.194(0.002) 0.624(0.031) 32.1%(0.003) 0.206(0.002) 0.664(0.137) 37.1%(0.005)