摘要
利用计算机程序对观察数据组教小于性状数这种结构类型的数据进行逐步回归是不稳定的。主要原因是计算时,入选和剔除性状的自由度计算越界,作者经过反复修改程序及运算表明:原有的程序通过调整入选和剔除性状的标准不能有效地控制入选性状的数目,但在按一般α水平进行逐步回归的基础上依据入选性状贡献率的大小再次选择,达到控制入选性状数目的方法是合理和正确的。用此方法计算多份实验资料结果表明,其复相关系数都在0.90以上。
While the observed datum's number is less than the charactert'snumber the results of stepwise regression using program are not stable because ofexceeding the limit of D.F. to set regression equation using computer program.It can't control the number of selected characters to adjust the criterion of selec-ted and delected characters, but it's rational and correct to control the numbersof selected characters through selecting them again according to the contributionratio of selected characters based on the results of stepwise regression in the lowerlevel of ɑ. The experiments show that the ratios of correct forcast are over 95%.
基金
国家自然科学基金资助项目系列论文之一
关键词
生物统计
逐步回归
程序设计
biostatistics
stepwise regressions
programming/abnomal datum matrix
