Skip to main content
. Author manuscript; available in PMC: 2014 Jun 5.
Published in final edited form as: Popul Stud (Camb). 2011 Dec 13;66(1):1–28. doi: 10.1080/00324728.2011.611411

Table 4.

Root-mean-squared-errors (RMSEs) for e0, 1q0, 45 q15, and 20q60 by sex, various model life table methods, Human Mortality Database life tables (n = 719)

Female Male
e0 1q0 45q15 20q60 e0 1q0 45q15 20q60
Given 5q0 only:
 Log-linear (bi-weight) 1.64 0.011 0.032 0.046 2.62 0.012 0.064 0.057
 Log-linear (OLS) 1.67 0.010 0.033 0.047 2.60 0.011 0.062 0.057
 Log-quadratic (bi-weight) 1.62 0.010 0.032 0.045 2.55 0.011 0.062 0.056
 Log-quadratic (OLS) 1.62 0.010 0.032 0.045 2.52 0.011 0.060 0.056
 Modified logit (re-est.) 1.66 0.010 0.032 0.048 2.47 0.011 0.060 0.052
 Modified logit (orig.) 1.85 0.015 0.034 0.051 2.56 0.019 0.067 0.053
 Coale-Demeny West Model 2.73 0.010 0.042 0.062 4.09 0.011 0.086 0.088
 UN General Model 4.50 0.011 0.070 0.110 5.67 0.010 0.104 0.139
Given 5q0 and 45q15:
 Log-linear (bi-weight) 0.83 0.011 0 0.047 0.69 0.012 0 0.045
 Log-linear (OLS) 0.78 0.010 0 0.046 0.62 0.011 0 0.045
 Log-quadratic (bi-weight) 0.70 0.010 0 0.042 0.59 0.011 0 0.041
 Log-quadratic (OLS) 0.70 0.010 0 0.042 0.57 0.011 0 0.041
 Modified logit (re-est.) 0.69 0.010 0 0.042 0.61 0.011 0 0.043
 Modified logit (orig.) 0.88 0.014 0 0.045 0.99 0.020 0 0.044
Best family given 5q0:
 Log-quadratic (5 families) 0.94 0.010 0.013 0.042 1.18 0.011 0.027 0.044
 Coale-Demeny (4 families) 2.45 0.010 0.026 0.061 3.90 0.014 0.077 0.084
 UN tables (5 families) 3.26 0.013 0.030 0.084 3.48 0.011 0.066 0.080
 C-D or UN (9 families) 2.41 0.011 0.023 0.062 3.39 0.013 0.064 0.077
Best family given 45q15:
 Log-quadratic (5 families) 0.99 0.013 0 0.041 1.61 0.017 0 0.042
 Coale-Demeny (4 families) 1.44 0.014 0 0.052 3.12 0.032 0 0.051
 UN tables (5 families) 1.37 0.016 0 0.047 1.92 0.018 0 0.062
 C-D or UN (9 families) 1.37 0.013 0 0.053 1.79 0.019 0 0.062
Notes:
  1. For these comparisons, the log-quadratic model was estimated using either ordinary least squares (OLS) or weighted least squares using a bi-square weight function of residuals: the bi-weight method. See Appendix for more explanation.
  2. Estimation errors for the log-quadratic model in the two sets of “best family” comparisons were derived using the model as estimated by the bi-weight method.

Source: As for Table 3