以下ml check的检查结果怎么看?ml max总是计算不出来

是 否 +2 论坛币

k人 参与回答

经管之家送您一份

应届毕业生专属福利!

求职就业群

赵安豆老师微信：zhaoandou666

经管之家联合CDA

送您一个全额奖学金名额~ !

立即领取

感谢您参与论坛问题回答

经管之家送您两个论坛币！

+2 论坛币

. ml check
Test 1:  Calling mynormal to check if it computes log likelihood and
         does not alter coefficient vector...
         Passed.
Test 2:  Calling mynormal again to check if the same log likelihood value
         is returned...
         Passed.
------------------------------------------------------------------------------
The initial values are not feasible.  This may be because the initial values
have been chosen poorly or because there is an error in mynormal and it
always returns missing no matter what the parameter values.
Stata is going to use ml search to find a feasible set of initial values.
If mynormal is broken, this will not work and you will have to press Break
to make ml search stop.
Searching...
initial:       log likelihood =     -<inf>  (could not be evaluated)
searching for feasible values ...+
feasible:      log likelihood = -6959.1055
improving initial values ..........
improve:       log likelihood = -6959.1055
rescaling entire vector +.
rescale:       log likelihood =   -5851.64
rescaling equations +++++.....
rescaling equations ......
rescale eq:    log likelihood = -4990.9165
restarting tests...
------------------------------------------------------------------------------
Test 1:  Calling mynormal to check if it computes log likelihood and
         does not alter coefficient vector...
         Passed.
Test 2:  Calling mynormal again to check if the same log likelihood value
         is returned...
         Passed.
Test 3:  Calling mynormal to check if 1st derivatives are computed...
         test not relevant for method lf.
Test 4:  Calling mynormal again to check if the same 1st derivatives are
         returned...
         test not relevant for method lf.
Test 5:  Calling mynormal to check if 2nd derivatives are computed...
         test not relevant for method lf.
Test 6:  Calling mynormal again to check if the same 2nd derivatives are
         returned...
         test not relevant for method lf.
------------------------------------------------------------------------------
Searching for alternate values for the coefficient vector to verify that
mynormal returns different results when fed a different coefficient vector:
Searching...
initial:       log likelihood =     -<inf>  (could not be evaluated)
searching for feasible values .+
feasible:      log likelihood = -6554.2216
improving initial values ..........
improve:       log likelihood = -6554.2216
continuing with tests...
------------------------------------------------------------------------------
Test 7:  Calling mynormal to check log likelihood at the new values...
         Passed.
Test 8:  Calling mynormal requesting 1st derivatives at the new values...
         test not relevant for method lf.
Test 9:  Calling mynormal requesting 2nd derivatives at the new values...
         test not relevant for method lf.
------------------------------------------------------------------------------
                         mynormal HAS PASSED ALL TESTS
------------------------------------------------------------------------------
Test 10: Does mynormal produce unanticipated output?
         This is a minor issue.  Stata has been running mynormal with all
         output suppressed.  This time Stata will not suppress the output.
         If you see any unanticipated output, you need to place quietly in
         front of some of the commands in mynormal.
-------------------------------------------------------------- begin execution
---------------------------------------------------------------- end execution



. ml max
initial:       log likelihood = -6554.2216
rescale:       log likelihood = -5437.9889
rescale eq:    log likelihood = -5059.2758
could not calculate numerical derivatives
flat or discontinuous region encountered
r(430);



还有那个initial 值是什么意思呢?一定要定义吗?

扫码加我 拉你入群

请注明：姓名-公司-职位

以便审核进群资格，未注明则拒绝

分享0 收藏0 回帖

关键词：check Chec HEC max derivatives 结果 检查 check max

以我的经验，这个可能是模型设定的问题，比如你用极大似然法估计线性模型y=Xb的时候，如果X中存在共线性，那么在数值计算LLF的时候，它的二阶导数也就是Hessian矩阵就不是奇异的。而stata里面的一些最优化算法，比如NR迭代方法公式是X(n+1)=X(n)-(H(LLF))^(-1)*(LLF)'，如果Hessian矩阵非奇异，那么(H(LLF))^(-1)求不出来，也就是could not calculate numerical derivatives
flat or discontinuous region encountered
再检查一下模型设定吧。

好的,谢谢啦,我再改改我的模型看看

呵呵,还有一个问题
就是ml max要计算多少次才会停止呢
我的怎么运行了1000次以上还不停止
每运行一次,Iteration log likelihood后面就出现backed up
但是就停不下来
有没有什么办法可以设定它计算的次数呢?

yuhepi 发表于 2010-1-11 00:01
呵呵,还有一个问题
就是ml max要计算多少次才会停止呢
我的怎么运行了1000次以上还不停止

请问解决了吗 我也遇到了一样的问题