yuzeng 看到很多软件计算出来的判别函数中都存在常数项,但R中的判别分析怎么没找到这一常数项? Call: lda(as.matrix(d1[, 1:7]), grouping = d1[, 8]) Prior probabilities of groups: 1 2 3 4 0.23076923 0.33333333 0.07692308 0.35897436 Group means: trait1 trait2 trait3 trait4 trait5 trait6 trait7 1 86.94667 12.472222 3.734444 2.003333 0.8522222 1.726667 0.1966667 2 86.93385 12.535385 3.429231 1.600000 1.4138462 1.835385 0.2861538 3 90.01667 23.890000 4.133333 3.550000 1.0166667 1.820000 0.9833333 4 89.59000 3.580714 4.915000 2.190714 0.9321429 2.159286 0.0450000 Coefficients of linear discriminants: LD1 LD2 LD3 trait1 0.03439346 -0.05245817 -0.02820474 trait2 -0.11958720 0.01416337 0.03413679 trait3 0.24268541 0.78901373 0.50185402 trait4 -0.60989882 0.63882476 -0.01899702 trait5 -0.52898138 0.40392250 -0.69295288 trait6 -0.71479449 1.27297461 -1.85357785 trait7 -0.81890540 0.71836209 -0.77801911 Proportion of trace: LD1 LD2 LD3 0.6331 0.3159 0.0510 这里怎么找不到判别函数的常数项?
yuzeng 谢谢搂住的关注,因为我看文献中判别分析所给出的函数都包括各项系数及常数项,同时在网上看到说在“When the data are standardized or are deviations from the mean”,此情况下才没有常数项,我不知道怎么回事,能不能请搂住解答下,谢谢了。 网页地址:http://faculty.chass.ncsu.edu/garson/PA765/discrim.htm
yuzeng When does the discriminant function have no constant term? When the data are standardized or are deviations from the mean. 上面是原网页中的内容,我的标准化是对原始数据进行标准化处理。