[R-sig-ME] [FORGED] Re: Question about gls --- whoops!!!
Louise.M.Ryan at uts.edu.au
Fri Mar 23 23:57:50 CET 2018
Yes, it is very weird. I tried all sorts of things and many gave me that strange second number.
But the good news is that I now have a solution that works!! Thank you so much :)
Sent from my iPhone
> On 24 Mar 2018, at 8:30 am, Rolf Turner <r.turner at auckland.ac.nz> wrote:
>> On 24/03/18 10:52, Rolf Turner wrote:
>>> On 24/03/18 08:58, Andrzej Galecki wrote:
>>> Hi Louise,
>>> One possible way ...
>>> mSt <- fit.gls$modelStruct
>>> cSt <- mSt$corStruct
>>> coef(cSt, unconstrained = FALSE)
>>> # Phi
>>> # 0.1748786
>> Not exactly obvious, but, is it? How on earth would one ever figure this out, except by appealing to r-sig-mixed-models?
>> Looking at str(fit.gls) one sees that there is a (list) component "modelStruct" and that this component in turn has a component "corStruct". If one is an optimist, one might try
>> which (mirabile dictu!) gives:
>> Correlation structure of class corAR1 representing
>> Again if one is an optimist, one might try stripping away all the unwanted baggage by doing
>>  0.3533897
> Nope! ***NOT*** bingo!!! Louise just pointed out to me that what I got by applying as.vector(), i.e. 0.3533897, is not the same as the number one gets by just typing fit.gls$modelStruct$corStruct, which is 0.1748786 --- which is what Andrzej's approach gives (and which would
> appear to be the right answer). Bit of a duhhh on my part.
> Setting unconstrained=FALSE as in Andrzej's approach seems to be crucial. I missed that bit ....
> But it still mystifies me why as.vector() doesn't give the same numeric value as is obtained from just printing the object in question. Also
> looking at the code of nlme:::coef.corStruct I would have thought that
> this function would simply throw an error when unconstrained is set to FALSE. (Just type "nlme:::coef.corStruct" and you'll see what I mean.)
> If I set foo <- nlme:::coef.corStruct and then do
> I do indeed get an error. WTF?
> There would appear to be no way for the results to be different (look
> at the code) --- but different they are.
> But that sort of nonsense is to be expected when S4 classes and methods are involved in any way.
> Technical Editor ANZJS
> Department of Statistics
> University of Auckland
> Phone: +64-9-373-7599 ext. 88276
ompanying attachments may contain confidential information. If you are not the intended recipient, do not read, use, disseminate, distribute or copy this message or attachments. If you have received this message in error, please notify the sender immediately and delete this message. Any views expressed in this message are those of the individual sender, except where the sender expressly, and with authority, states them to be the views of the University of Technology Sydney. Before opening any attachments, please check them for viruses and defects. Think. Green. Do. Please consider the environment before printing this email.
More information about the R-sig-mixed-models