Skip to main content
. Author manuscript; available in PMC: 2011 Jan 1.
Published in final edited form as: Cancer Res. 2010 Jan 12;70(2):552–559. doi: 10.1158/0008-5472.CAN-09-2653

Table 3.

Summary of validation results.

Concordance
index C*
Observed/Expected
O/E ratio
Mean
Squared
Error
MelaPRO-HBI 0.86 (0.75, 0.97) 0.85 (0.62, 1.08) 0.06
(0.03, 0.08)
MelaPRO-LBI 0.86 (0.74, 0.97) 0.31 (0.20, 0.42) 0.19
(0.15, 0.22)
MelaPRO-GEM 0.83 (0.67, 0.98) 0.54 (0.38, 0.72) 0.08
(0.06, 0.11)
MELPREDICT 0.82 (0.61, 0.93) 1.10 (0.92, 1.17) 0.09
(0.04, 0.12)
Difference Between
MelaPRO-HBI and
MelaPRO-LBI
0.007
(−0.01,0.02)
Not Applicable§ −0.13
(−0.16, −0.10)
Difference Between
MelaPRO-HBI
and MelaPRO-GEM
0.04
(−0.008, 0.09)
Not Applicable −0.03
(−0.04, −0.02)
Difference Between
MelaPRO-HBI and
MELPREDICT
0.05
(0.007, 0.17)
Not Applicable −0.03
(−0.06, 0.002)
Sensitivity Specificity
FH:≥2 affected relatives 0.81 0.58
*

The concordance index C is equal to the AUC (area under the ROC curve)

The ratio between the observed number of carriers and the total number of predicted carriers

Leave-one-out cross-validation on Boston data using fixed covariates for MELPREDICT

§

Computing the difference in this case is not appropriate, as each ratio should be compared to the reference value of one.