Introduction systematic, environmental and genetic effects direct-

254 M . Serrano et al. Livestock Production Science 67 2001 253 –264

1. Introduction systematic, environmental and genetic effects direct-

ly, where they are expressed, on the day of record- In Latxa and Manchega Spanish dairy sheep, test ing. day record TD collecting is carried out under the Effects that influence TD production are not International Rules for Milk Sheep Recording different from those that affect complete lactation ICAR. Most records in both breeds are collected yield in general Danell, 1990. Nevertheless, for TD monthly under an alternate morning evening system measures there are differences from lactation records following ICAR rules. To standardize records to 120 and, therefore, different approaches have been used days, at least four TD per ewe are required, other- to consider environmental effects. Modeling the herd wise, in Manchega ewes, lactation is projected to 120 effect as herd-test day HTD instead of herd-year- days by extension factors developed previously season of parturition HYS leads to reduced residual Serrano et al., 1996a. In Manchega, 16 of the variances Ptak and Schaeffer, 1992, 1993. Includ- available lactation records lack the third and or four ing new environmental effects, such as the stage of TD and need to be projected. When the first and or lactation measured in number of days in milk, allows second TD records are missing, lactations are not a better estimation of environmental variability. projected and the records are discarded 3. However, modeling TD records causes some The goodness of the standardization and projec- problems with respect to the contemporary group tion procedures depends on the quality of milk CG definition, the treatment of within and among recording with regard to temporal aspects. TD re- parity information and the increase of computational cords are frequently collected at highly variable time demands due to the larger number of records to be periods due to animal management. This implies that managed. estimation of standardized yields depends not only Various authors Meyer et al., 1989; Ptak and on TD yield but also on the lactation stage of the Schaeffer, 1992, 1993; Stanton et al., 1992; Strabel animal from which test samples were collected. This and Szwaczkowski, 1995; Swalve, 1995; Rekaya, traditional approach of using lactation records has 1997 have suggested that models with HTD classes been criticized as inconsistent, since the aggregation are superior to HYS based on residual variances. of records taken at defined locations and time is However, the HTD model is rather problematic rather trivial. regarding the size of the CG. In several studies Numerous studies in dairy cattle Danell, 1982; Reents et al., 1995; Swalve, 1995; Van Bebber et al., Meyer et al., 1989; Pander et al., 1992; Ptak and 1997 it has been pointed out that this may be an Schaeffer, 1993; Reents et al., 1995; Swalve, 1995; important obstacle when trying to fit a HTD model. Rekaya, 1997; Wiggans and Goddard, 1997 and In dairy cattle, different models have been pro- fewer in goats Schaeffer and Sullivan, 1994 and posed to estimate the covariance structure among sheep Baro et al., 1994; Barrillet and Boichard, TD measures. Repeatability models that consider TD 1994; Georgoudis et al., 1997; El-Saied et al., 1998; as repeated measures of the same trait within lacta- Serrano et al., 1998 have dealt with the use of TD tion and as multiple traits across lactations Reents et records as an alternative to standardized lactation al., 1995. Models where TD are considered as yields. Use of test days has the advantage of directly different traits within and along lactations Wiggans considering records at their origin. Under TD and Goddard, 1997, for which the main constraint is models, records on the sample day are considered the high computational demand. Random regression directly in analysis and no assumption about the models RRM are also being applied to fit the shape length of a lactation has to be made Visscher and of the lactation curve Schaeffer and Dekkers, 1994; Goddard, 1995. Models that consider single TD Jamrozik and Schaeffer, 1997; Rekaya et al., 1999. records allow the removal of abnormal measures This kind of model seems to be adequate for traits avoiding the elimination of entire lactations. On the which repeated measurements are taken on a trajec- other hand, records do not have to be projected using tory of time and allows us to obtain information on a ‘extension’ factors and all information can be used. trait that changes gradually over time. Finally, the With TD models an attempt is made to account for use of covariance functions to fit different covar- M . Serrano et al. Livestock Production Science 67 2001 253 –264 255 iances among repeated records has been proposed yield and protein percentage TD records. Given the Kirpatrick et al., 1994; Meyer and Hill, 1997 to large size of the data set available 94,191 lactation reduce the number of parameters to fit when there records and 70,207 pedigree records, several edits are many measures for each individual taken over were performed so that the data set was more time which are considered as different traits. manageable and a better correspondence between test In this study, a genetic evaluation model is defined day number and date of test was achieved. First TD and genetic parameters are obtained for TD milk between days 30 and 72 from parturition and time yield and protein percentage records in two different interval between successive tests of more than 25 Spanish dairy sheep breeds. Some considerations are days and less than 35 days were forced. In addition, made to develop the models. Firstly, since in dairy a minimum milk yield in each test of 200 ml and no ewes there are generally four TD records in a missing values for any TD were required. Identifica- lactation in contrast to the 10 in dairy cattle, compu- tion of dam and number of lambs born plus existence tational demands are much smaller and allow a of both milk and protein percentage records were multivariate approach within lactation. Secondly, also required. Finally, data were sampled by HYS dairy sheep herds have a larger average size than CG, discarding CG with less than six observations. dairy cattle herds, which partially avoids problems After these edits, 22,804 milk yield and protein with HTD CG size. percentage TD records from 5701 first lactation ewes of 32 herds were kept and used to estimate genetic parameters. The pedigree file consisted of 10,883

2. Material and methods animals.