Abstract
A modified regression-type estimator has been constructed for two-phase sampling using two auxiliary variables w, x and a measure of size z. The estimator has been developed by using probability proportional to a measure of size with replacement at first phase and equal probability sampling without replacement at the second phase. The mean square error of the modified estimator has been derived under arbitrary probabilities of selection. An empirical study has been made of the performance of new estimator with Roy's estimator (2003).