Testing of the Prediction Unbiasedness on the Basis of Janus Quotient
Testing of the Prediction Unbiasedness on the Basis of Janus Quotient

Summary/Abstract: The problem of choosing the appropriate predictor is being considered. Generally, in this paper the analysis is focused on the problem of unbiasedness of the predictors. Several tests attempting to verify the unbiasedness of three predictors of the linear trend are proposed. They are based on some modifications of the well-known Janus quotient being a ratio of the variance of prediction errors and the residual variance. In general each of the considered test statistic can be represented as the ratio of two quadratic forms of normal vectors. These two quadratic forms can be dependent, so its distribution function has to be approximated. An example of testing hypothesis on unbiasedness is presented. The obtained results can be generalized in the case of prediction on the basis of regression models.

• Details
• Contents