※ 引述《Yyh (我爱批踢踢)》之铭言：
: : I think I may have misunderstood how SVD works. (?)
: I don't see any mistake in your code.@@"
: What dataset you used in above result?
: lm is lsfit?
: if yes, I don't remember I have result with "NA" in any dataset.
Thanks for your response. :)
I am testing on dataset "house", and I think the reason I got "NA"
is because I use only first 20 instances for test.
When I use all data in the house, the attribute I get from SVD is the
same with that from lsfit.
I am still trying to find out why the number of instance
make a difference in calculating coefficient by SVD and by lsfit.