Various methods of simulating lactation milk yield from regular and irregular monthly, bimonthly and trimonthly records were evaluated on individual daily milk weights from 1296 lactations of half-bred dairy cattle of the Department of Animal Breeding, Haryana Agricultural University, Hisar, India. Simulation method 1 was heavily biased with larger variations in starting date i.e. the day of first test. Simulation method 2 was clearly superior to method 1 but was generally not suitable. Simulation method 3, involving adjustment for actual length of interval between consecutive samples, resulted in the smallest error for regular monthly, bimonthly and irregular monthly intervals with starting date between days 10 and 45. Regular trimonthly and irregular bimonthly sampling resulted in less average error (within a range of ± 3%) for starting dates between days 10 and 30 as compared with others. However, no simulation method was satisfactory for irregular trimonthly sampling.