1# kelubeilesi

回归分析中,回归系数是根据数据拟合出来的估计值,你没有办法去限制其小于多少,或者大于多少,但你可以检验系数是否等于多少.所以你只能去检验a=某个常数,或者a=b=0 或者a+b=1

貌似1stopt应该可以。其他一些软件也应该可以的吧

The answer is positive. A regression problem is not thing but an optimization problem if you think that way. The restrictions are equivalent to constrains in optimization problems.

Some procedures handle them as bounds.

可以用proc nlin过程的bounds statement进行限制。

I am agree with bobguy. For this problem what you really need to do is to minimize the ||Y- X beta_hat||^2 subject to the constrain of a+b<1.  It is quite easy to program.