[R] StructTS standard errors

Felix Andrews felix at nfrac.org
Tue Dec 22 23:00:48 CET 2009


Hi David,

You might not be aware that Giovanni Petris is the author of the dlm
package, a world-class package for modelling stochastic / time-varying
systems.

Not that I know much about this stuff.

Cheers
-Felix


2009/12/23 David Winsemius <dwinsemius at comcast.net>:
> On Dec 22, 2009, at 11:55 AM, Giovanni Petris wrote:
>
>>
>> David,
>>
>> Thank you for the insight.
>
> Right.
>>
>> HA1) StructTS() ships with R in package stats.
>>
>> HA2) I don't see anything strange or unusual in looking for standard
>> errors
>>        of estimated parameters. In structural time series the only unknown
>>        parameters, estimated by StructTS(), happen to be variances.
>>        (Incidentally, I know how to take square roots in R)
>
> Yes. I figured most of this out after doing a bit of searching. Both of my
> off-target points amply illustrate my lack of experience with this
> particular time-series function.
>
> So let me ask a question: Would the domain of stochastic processes be
> relevant here? I am wondering if the theoretical results in Ito calculus
> might let you assign a more reasonable estimate for future behavior. My
> memory is that random processes with measured (past) variances under
> conditions of "temporal homogeneity" (there's probably a more mathematically
> correct terminology)  should have a predictable relationship to the
> underlying variance parameter.
>
> --
> David Winsemius
>
>>
>> Best,
>> Giovanni
>>
>> ----- Original Message -----
>> From: David Winsemius <dwinsemius at comcast.net>
>> Date: Monday, December 21, 2009 3:50 pm
>> Subject: Re: [R] StructTS standard errors
>>
>>> Alternate hypotheses:
>>>
>>> HA1: There is no one in the readership who is familiar with the
>>> function (unlikely), or among those lacking knowledge of what
>>> package
>>> it might reside in (such as myself)  chose not to look up what you
>>> might have provided on the first posting: a self-contained example
>>> to
>>> work with.
>>>
>>> HA2: Your question also suggests a certain statistical confusion.
>>> Do
>>> your really want standard errors of estimated variances? That would
>>> be
>>> somewhat unusual. Or do you want to take the square-roots of the
>>> variances to get standard errors of the estimated coefficients?
>>>
>>>
>>> On Dec 21, 2009, at 4:26 PM, Giovanni Petris wrote:
>>>
>>>>
>>>> Sorry for asking again, but I did not receive any answers - I
>>>
>>> know
>>>>
>>>> it's a busy time... ;-)
>>>>
>>>> Shell I conclude that it is not possible to obtain standard
>>>
>>> errors
>>>>
>>>> from StructTS()
>>>> or its output?
>>>>
>>>> Thanks,
>>>> Giovanni Petris
>>>>
>>>> ----- Original Message -----
>>>> From: Giovanni Petris <gpetris at uark.edu>
>>>> Date: Thursday, December 17, 2009 10:13 am
>>>> Subject: [R] StructTS standard errors
>>>>
>>>>>
>>>>> Hello,
>>>>>
>>>>> Does anybody know if (and how) it is possible to obtain standard
>>>>> errors of estimated variances from StructTS? (R 2.10.0).
>>>>>
>>>>> Thank you in advance,
>>>>> Giovanni
>>>
>>> --
>>> David Winsemius, MD
>>> Heritage Laboratories
>>> West Hartford, CT
>>>
>>>
>
> David Winsemius, MD
> Heritage Laboratories
> West Hartford, CT
>
> ______________________________________________
> R-help at r-project.org mailing list
> https://stat.ethz.ch/mailman/listinfo/r-help
> PLEASE do read the posting guide http://www.R-project.org/posting-guide.html
> and provide commented, minimal, self-contained, reproducible code.
>



-- 
Felix Andrews / 安福立
Postdoctoral Fellow
Integrated Catchment Assessment and Management (iCAM) Centre
Fenner School of Environment and Society [Bldg 48a]
The Australian National University
Canberra ACT 0200 Australia
M: +61 410 400 963
T: + 61 2 6125 4670
E: felix.andrews at anu.edu.au
CRICOS Provider No. 00120C
-- 
http://www.neurofractal.org/felix/




More information about the R-help mailing list