Effects of long-time series of data on genetic evaluations for performance of Swedish Warmblood riding horses.

For Swedish Warmblood sport horses, breeding values (BVs) are predicted using a multiple-trait animal model with results from competitions and young horse performance tests. Data go back to the beginning of the 1970s, and earlier studies have indicated that some of the recorded traits have changed through the years. The objective of this study was to investigate the effects of including all performance data or excluding the older ones compared to a bivariate model (BM) considering performance traits in early and late periods as separate traits. The bivariate approach was assumed to give the most correct BVs for the actual breeding population. Competition results in dressage and show jumping for almost 40 000 horses until 2006 were available. For riding horse quality test (RHQT), data of 14 000 horses judged between 1973 and 2007 were used. Genetic correlations of 0.69 to 1.00 were estimated between traits recorded at different time periods (RHQT data) or different birth year groups (competition data). A cross-validation study and comparison of BVs using different sets of data showed that most accurate and similar results were obtained when BVs were predicted from either the BM or the univariate model including all data from the beginning of the recording. We recommend using all data and applying the univariate model to minimise the computational efforts for genetic evaluations and for provision of reliable BVs for as many horses as possible.