In the present study, the prognostic values of multiple and spline regression models were tested for 305-day lactation milk yield of cows. The predictors were: HF genes proportion in cow's genotype; average milk yield for 305-day lactation from 4 first milkings of all cows in a barn, in which a cow was used in a given year; month of calving; and average daily milk yield from first four test-day milkings. Models were developed basing on 628 first lactations of BW cows with average 71% HF genes proportion. Subsequently, the predictive values of the models examined were verified on the grounds of next 105 first lactations. Prognostic differences of the models examined were determined, finding the prognosis obtained with the spline regression more accurate (smaller prediction error, higher coefficient of correlation for prognosis and real values in the model containing information from first three test-day milkings). These models are easy to construct and may be useful in practical estimation of cow lactation yields. They may be used for predicting the actual lactation yield in order to minimise production costs and achieve better production results.