Construction Formula of Biological Age Using the Principal Component Analysis
Table 5
Comparisons between the multiple linear regression (MLR), the principal component analysis (PCA), Hochschild’s method, and the Klemera and Doubal method (KDM) in biological age (BA) estimates.
Methods
Advantages
Disadvantages
MLR
As an initial method of BA estimates, MLR can detect the stabilization and multicollinearity of the empirical data [26].
The distortion of BA at the regression edge is influenced by mathematical factors, and MLR also ignores the discontinuity of the ageing rate during the whole life of individuals [9, 27, 28]. The BA formula established by MLR is probably disserviceable [8].
PCA
PCA selects and transforms the original biomarkers to a reduced and/or transformed new series of uncorrelated variables [8]. PCA avoids the influence of regression edge in MLR [28] and is easy to be operated. This method generates the uncorrelated variables and provides the information of underlying structure of variables [26].
The final step of the computation resembles the MLR method, and some of the statistical deficiencies of MLR cannot be totally avoided [8].
Hochschild’s method
Hochschild’s method uses the regression for individual biomarkers and evaluates the biomarkers according to their impact on life expectancy [8].
Hochschild’s method is not based on mathematical definition of BA, and the construction mechanism is elusive. In particular, the calculation does not correspond to the optimum algorithm [8]. Moreover, a large number of subjects are needed to be measured when Hochschild’s method is adopted for a newly developed system [26].
KDM
KDM is a more reliable predictor of mortality and performed better than the chronological age [16]. KDM gives lower errors than other methods, evaluates the precision of BA estimates, and solves the paradox of biomarker selection according to CA [8].
