Re: Deat Jun,the covariance?
Dear Hongpo,
In my opinion, we develop a model specific for a data set. If your data set
supported estimation of such a covariance then I think you should include it
(as indicated by a statistical significant drop of OFV). It is better if you
could find some physiological meaning for those off diagonal covariances (as
mentioned by Saik). But it is not a reason enough to exclude a covariance
just because you couldn't find a physiological relevance. On the other hand,
even if you have a strong belief that a covariance exists, if your data set
doesn't support, I really doubt inclusion of such an unidentifiable
covarinace would do any good unless you have good prior knowledge (then you
can try to fix it).
Welcome more comments.
Jun
Millipore Corporation
Quoted reply history
2009/12/14 yhb5442387 <[email protected]>
>
> Dear Jun:
> Sorry for making you in the reading trouble:
> Here is my question:
> Is there any special specification on the employ of covariance:
> The papers which I have read most always take the covariance excluded,such
> as W21,w31 w32.If the Decrease of OFV between the model of the covriance
> included and the one not is significantly meaningful,what choice should I
> do? I really do not understand the correlation of Cl and V,if the
> convariance between the two parameters does exist.Dose it truly exist
> in physiologically or just mathematically?
>
>
> Dear Hongpo,
>
> I am not sure I understand your question. Can you rephrase?
>
> Jun Shen
> Millipore Corporation
>
> 2009/12/13 yhb5442387 <[email protected]>
>
>>
>> Dear professor:
>>
>> THE off diagnoal:
>>
>> Sometimes we make the OMEGA syntax as follows:$omega 0.1 0.1,but $omega
>> blokc (2) 0.1 0.01 0.1 at another occasion.As we know,the only difference is
>> whether the off diagonal estimate be used.So is there any specila condition
>> about the existence of the W21,or some another W square like that?
>>
>> Thank you !!!
>>
>
> --
> 工作和生活,都要开心的过.
> 你好,叶红波在此送上真挚的祝福.祝你开心,
> 叶红波
>
>
>