An estimation method for an unknown covariance in cross-section adjustment based on unbiased and consistent estimator

Abstract

ABSTRACT A new estimation method of an unknown covariance in the cross-section adjustment method for the development of an application library is proposed. The unknown covariance is defined by the difference between the true covariance (the population covariance) and a prior covariance assumed by an analyst. The unknown covariance is estimated using an empirical covariance consistent with the observed data. To estimate the unknown covariance, an unbiased and consistent estimator in regression analysis has been incorporated into the conventional cross-section adjustment. This estimator does not require assumptions for the probability distribution of the observation data. The statistical properties of this estimator were numerically verified. In addition, the effectiveness of the proposed method was confirmed by another numerical test using actual integral experimental data. In the second numerical test, the modeling uncertainty (covariance) due to the deterministic analysis method was assumed to be unknown. The results show that the proposed method can practically estimate the unknown covariance and adjusted cross-sections using only prior information on covariances.