The data consisted of 3,740 Holstein lactations at six dairy farms from 1969 to 1984 in Korea. In order to estimate the extension factor the S78 incomplete records were deleted and 3,162 records were only used. The generalized least squares analysis for milk yield used a model including mean, herd, year, season and parity as fixed effects. In order to investigate accuracy for each extension factors, variance of difference. between actual and estimated yield were estimated. The effects of herd, calving year, calving month and parity for milk yield were highly significant (P$lt;0.01), and were mostly influenced at 8-10th, 2-4th and 1-3rd lactation stage respectively. In order of size of variance of difference between actual and estimated yield, method P showed the smallest variance, followed by the methods of multiple regression factor, ratio factor (estimated the remaining part of the lactation from the yield on the last test day), cumulative regression factor and cumulative ratio factor. When environmental effects such as herd, calving year, calving month or parity were not eliminated, variance generally became larger. The extension factor using ratio increased the variance of difference by about 2S% and that using regression increased the variance of difference from 38.61 to 48.0%.