Dimension-reduced cross-section adjustment method based on minimum variance unbiased estimation

Yokoyama, Kenji   ; Yamamoto, Akio*; Kitada, Takanori*

A new formulation of the cross-section adjustment methodology with the dimensionality reduction technique has been derived. This new formulation is proposed as the dimension reduced cross-section adjustment method (DRCA). Since the derivation of DRCA is based on the minimum variance unbiased estimation (MVUE), an assumption of normal distribution is not required. The result of DRCA depends on a user-defined matrix that determines the dimension reduced feature subspace. We have examine three variations of DRCA, namely DRCA1, DRCA2, and DRCA3. Mathematical investigation and numerical verification have revealed that DRCA2 is equivalent to the currently widely used cross-section adjustment method. Moreover, DRCA3 is found to be identical to the cross-section adjustment method based on MVUE, which has been proposed in the previous study.