Single regression (SR), multiple regression (MR), and factor analysis (FA) methods for projecting lactation milk yield from part-lactation test-day records at regular 15-, 30-, and 45-day intervals were evaluated on individual daily milk weights from 707 half-bred dairy cattle of the Department of Animal Breeding, CCS Haryana Agricultural University, Hisar, India. Both MR and FA used all known test-day yields, while SR used only the last known test-day yield. The SR method was most effective in estimation of yield with minimum average and standard error of absolute bias and distribution of absolute error. Taking the first test on the 15th day for the 15-day test interval and on the 30th day for the 30- and 45-day test intervals resulted in estimation of lactation milk yield with smaller absolute bias and standard error. Part lactation test-day records available up to 210 days was considered to be optimum as there was no significant improvement by utilising test-day records beyond that. Regular sampling at 45-day intervals with the first test on the 30th day post-partum and subsequent test day records up to 210 days lactation for projecting lactation milk yield by the SR method was recommended as the most practical, simple, and efficient for field recording.