|（Department of Physical Education，Taizhou University，Linhai 317000，China）
Abstract: When such variable deletion methods as OLS stepwise regression are used to process collinear data in the sports training area, they will reject some important explanatory variables or sample points out of the models, thus weakening the theoretical priority position and guidance function. The PCR model is not well capable of integrating and screening system information, and its fitting and predicting precisions are not so ideal. The PLS is well capable of integrating and screening system information, and its fitting and predicting precisions are relatively high; it is currently a very effective statistical method for processing collinear data with a small sample and multiple indexes in the sports training area, and it provides a possibility for the empirical test of theoretical models of sports training science, but it has some difficulty to specifically identify the physical meaning of components abstracted.
Key words: sports training science；sports trainingof mathematical statistics；partial least-square regression；regression modeling