Comparison of the inverse estimator with the classical estimator subject to a preliminary test in linear calibration

Abstract The classical estimator in the linear calibration problem, obtained by inverting the regression relationship, has a great disadvantage viz., the mean and variance are unbounded. For overcoming this difficulty a practical truncation procedure based on a test of hypothesis about the regression coefficient is suggested. The properties of such a conditional estimator are studied and it is demonstrated that the suggested procedure removes the above mentioned disadvantage of the classical estimator. The exact expressions for mean and mean square error of an alternative estimator, known as the inverse estimator, have also been derived for comparison purposes.