t****a 发帖数: 144 | 1 已经系统估计出:
y=ax+bz
x=cm+dn
a, b, c, d都是consistent的
现在想知道m对y的影响,那么系数是a*c。
问题是:a*c的标准差怎么算?
谢谢! | t****g 发帖数: 715 | 2 My two cents:
(1) why not rewrite the model as y=ac*m+ad*n+bz? If you can do so, then
estimate ac directly and derive its variance afterwards;
(2) if rewriting the model is forbidden for some reason I can not see here,
then note that:estimator of a is normally distributed, as well as estimator of c, and their correlation can be consequently derived,
(2.1) apply delta method to get variance of ac, check your 1st year lecture notes if you forget what delta method is;
(2.2)a tedious way, but it may
【在 t****a 的大作中提到】![](/moin_static193/solenoid/img/up.png) : 已经系统估计出: : y=ax+bz : x=cm+dn : a, b, c, d都是consistent的 : 现在想知道m对y的影响,那么系数是a*c。 : 问题是:a*c的标准差怎么算? : 谢谢!
|
|