The selection criterion considered in the Latxa dairy sheep breeding scheme (blond‑faced and black‑faced ecotypes) for milk production is 120-days typified lactation. Using 1995‑1998 milk recording data, four alternative methodologies for predicting 120‑days typified lactation from part‑lactation were compared. Combination of fixed effects was considered through 48 different groups of comparison. These four methods were the following: first, the Modified Regression method extends from part to total‑lactation by linear regression on the last‑known test‑day record. Then, the other three methods use standard lactation curves of Wood, Ali‑Schaeffer and Morant‑Gnanasakthy. Four methods use different parameters according to the group of animals considered. All of them were compared with the current methodology, which uses simple regression for 1 known test‑day record and modified regression for 2 or more known test‑day records.
Although the fitting achieved with all these four methods can be considered to be good, that using Wood's equation was the best one according to different parameters. Knowing only 1 or 2 test‑day records, correlations between real data and predicted lactations were 0.95 and 0.99 respectively. Bias were regarded to be small (maximum of 0.22 litres).
Genetic correlations between both traits (real lactations and predicted lactations) were estimated to be around 0.99, showing a high accuracy in the genetic evaluation using predicted lactations.
In consequence, the Latxa sheep breeding scheme will change the current predicting lactation methodology to the one proposed here, which uses Wood's equation.