[R] [FORGED] Re: Compare two normal to one normal

[Mark Leeds]

Hi Rolf: I have  read  a decent amount about  the AIC  but that was a long,
long time ago. I too am no expert on it and John should read some of the
AIC literature John: There's one whole supposedly great text just on AIC
but I don't have it.  Link is here. Of course, it's
absurdly expensive but does get pretty good reviews.

http://www.amazon.com/Model-Selection-Multimodel-Inference-Information-Theoretic/dp/0387953647/ref=sr_1_1?ie=UTF8&qid=1443026829&sr=8-1&keywords=model+selection+aic

Note that if  you end up using the AIC approach, you'll still need the log
likelihoods in both models. I would calculate them yourself and all the
constants like 1/radical 2pi don't need to be included of course since
they'll just be scaling factors.











On Wed, Sep 23, 2015 at 2:22 AM, Rolf Turner <r.turner at auckland.ac.nz>
wrote:

> On 23/09/15 16:38, Mark Leeds wrote:
>
>> John: After I sent what I wrote, I read Rolf's intelligent response. I
>> didn't realize that
>> there are boundary issues so yes, he's correct and  my approach is EL
>> WRONGO. I feel very not good that I just sent that email being that it's
>> totally wrong. My apologies for noise
>> and thanks Rolf for the correct response.
>>
>> Oh,  thing that does still hold in my response is  the AIC approach unless
>> Rolf
>> tells us that it's not valid also. I don't see why it wouldn't be though
>> because you're
>> not doing a hypothesis test when you go the AIC route.
>>
>
> <SNIP>
>
> I am no expert on this, but I would be uneasy applying AIC to such
> problems without having a very close look at the literature on the
> subject.  I'm pretty sure that there *are* regularity conditions that must
> be satisfied in order that AIC should give you a "valid" basis for
> comparison of models.
>
> AIC has most appeal, and is mostly used (in my understanding) in settings
> where there is a multiplicity of models, whereby the multiple comparisons
> problem causes hypothesis testing to lose its appeal. Correspondingly AIC
> has little appeal in a setting in which a single hypothesis test is
> applicable.
>
> I could be wrong about this; as I said, I am no expert.  Perhaps younger
> and wiser heads will chime in and correct me.
>
> cheers,
>
> Rolf
>
>
> --
> Technical Editor ANZJS
> Department of Statistics
> University of Auckland
> Phone: +64-9-373-7599 ext. 88276
>

	[[alternative HTML version deleted]]