USING TRANSFORMATION ALGORITHMS TO ESTIMATE (CO)VARIANCE COMPONENTS BY REML IN MODELS WITH EQUAL DESIGN MATRICES

The reduction of computational demands on Restricted Maximum Likelihoo d (REML) procedures by a diagonalization approach is extended to multi ple traits by the use of canonical transformations. A computing strate gy is developed for use on large data sets employing two different REM L algorithms for the estimation of (co)variance components. Results fr om a simulation study indicate that (co)variance components can be est imated efficiently at a low cost on a wide range of computer systems.