Skip to main content
. Author manuscript; available in PMC: 2016 Feb 5.
Published in final edited form as: Psychol Aging. 2015 Sep 21;30(4):911–929. doi: 10.1037/pag0000046

Table 4.

Average Relative Efficiency Across Parameters, Simulation A

Listwise deletion Log weights CART weights RF weights MI
30% 50% 30% 50% 30% 50% 30% 50% 30% 50%
N = 100
 Linear   .955 .886   .954   .958 1.006 1.053   .994 1.075   .942 .885
 One split   .788 .896   .893   .959 1.010 1.095 1.001 1.197   .790 .887
 Two splits   .880 .821   .931   .982 1.016 1.049 1.032 1.168   .887 .822
 Three splits   .893 .884   .935   .943 1.018 1.073 1.006 1.081   .900 .881
N = 250
 Linear 1.006 .970 1.009 1.074 1.085 1.179 1.057 1.233 1.009 .971
 One split   .895 .971 1.069 1.072 1.152 1.167 1.331 1.348   .896 .976
 Two splits   .902 .902   .978 1.103 1.121 1.131 1.227 1.395   .910 .911
 Three splits   .945 .957 1.009   .998 1.135 1.171 1.102 1.191   .948 .946
N = 500
 Linear   .940 .917   .948 1.036 1.010 1.051 1.012 1.281   .937 .933
 One split   .816 .932 1.005 1.037 1.007 1.019 1.360 1.327   .828 .937
 Two splits   .883 .858   .955 1.042 1.052 1.019 1.208 1.507   .89 .872
 Three splits   .867 .873   .909   .899 1.040 1.009 1.076 1.276   .868 .883

Note. Relative efficiency is computed as the efficiency of measure X over the efficiency of pruned CART. Log = logistic regression; CART = classification and regression trees; RF = random forests; MI = multiple imputation.