为QTL作图的模型选择提供参考，基于标记与性状的多元回归建模，采用向前向后逐步方法对变量进行选择回归，着重于岭回归与偏最小二乘回归来解决变量间共线性问题并对BarleyDH.BIP群体进行建模分析。由于标记变量存在很强的共线性，在解决共线性问题中，仅仅对变量的选择剔除是不够，岭回归方法能建立相对较为精确的模型，但是预测不稳定，效果不够理想，偏最小二乘回归方法虽然模型建立没有岭回归精确，但是预测值与真实值最相近,可靠性较高。可根据QTL作图过程中不同需要进行各自合适的选择。

Provided the reference for the model selection of QTL(quantitative trait locus)mapping .Based on the markers and trait of multiple regression model ,using forward ,backward and stepwise methods to decide the choice of variable regression .Focuses on ridge regression and partial least-square regression to solute the linear problem between variables Since the strong altogether linear between the mark variables,it is not enough to eliminate some variables.Ｏn the one hand ,Ridge regression can establish relatively accurate model ,but it is unstable to predict and the effect is not good enough.Ｏn the other hand,although the model establishment of partial least-squares regression method is no better than ridge regression , the partial least-squares regression can get much closer prediction. We can choose appropriate method accorded to the different need for quantitative trait mapping.