Skip to main content

Table 4 Prediction errors from models to correct bias in indirect estimates of U5M

From: Measuring and correcting bias in indirect estimates of under-5 mortality in populations affected by HIV/AIDS: a simulation study

MethodSampleRoot mean square errorRoot median square errorMean relative errorMedian relative error
Full Linear Modelin-sample0.0029730.0011112.3400.390
out-of-sample0.0149220.0041831.9701.082
Forward Sel. BICin-sample0.0029880.0011432.3250.392
out-of-sample0.0149140.0041521.9661.087
Forward Sel. AICin-sample0.0029750.0011322.3240.389
out-of-sample0.0149250.0041841.9681.084
Backward Sel. BICin-sample0.0029740.0011102.3310.389
out-of-sample0.0149230.0041821.9661.082
Backward Sel. AICin-sample0.0029740.0011102.3310.389
out-of-sample0.0149230.0041821.9661.082
glmnet, alpha = 0in-sample0.0032410.0011352.5710.406
out-of-sample0.0032560.0011501.1000.382
glmnet, alpha = 0.5in-sample0.0029850.0011392.2650.389
out-of-sample0.0029680.0011770.9650.376
glmnet, alpha = 1in-sample0.0029850.0011372.3140.391
out-of-sample0.0029670.0011750.9600.374
PCR, ncomp = 20in-sample0.0039890.0020944.2950.647
out-of-sample0.0147000.0053513.0041.455
PCR, ncomp = 30in-sample0.0031220.0014301.7890.407
out-of-sample0.0149530.0045041.8681.101
PCR, ncomp = 35in-sample0.0029940.0011631.7220.369
out-of-sample0.0149250.0042811.8361.087
PLS, ncomp = 16in-sample0.0029880.0011541.8440.382
out-of-sample0.0149250.0042641.8721.087
PLS, ncomp = 32in-sample0.0029730.0011102.3560.389
out-of-sample0.0149220.0041831.9761.082
  1. Note: BIC Bayesian Information Criterion, AIC Akaike Information Criterion, glmnet generalized linear model via penalized maximum likelihood, where alpha is the elastic-net penalty term, PCR principle components regression; PLS partial least squares, ncomp number of components