In addition to the models that Simon mentions which are based on
state-space modeling, there is also an R-package (isn't there always!)
written by Paul Gilbert (Bank of Canada) called Time Series Factor
Analysis (tsfa) that seems more suited to multiple dimensions if that is
the problem you are working on. There are also psychometric models by
Molenaar and Nesselroade and some rudimentary code at:
http://www.psychstat.org/us/sort.php/26.htm%20-%20Dynamic%20Models
Best,
Greg
On Mon, June 8, 2009 6:06 pm, Simon Jackman wrote:
> IRT would be appropriate for binary or ordinal indicators. Recall
> that the vanilla IRT model is for cross-sectional binary data (e.g., "data
> from educational testing scored "right" or "wrong", or, in political
> science, recorded votes in deliberative bodies). A longitudinal version
> of the model can be developed where the latent variable has some kind of
> dynamic structure, that induces conditional independence among the
> indicators.
>
> The original post asks about factor analysis in a "repeated measures"
> setting. So I'll presume that we've got continuous indicators. If so, the
> appropriate model is a DLM (dynamic linear model), with West and Harrison
> being the canonical cite, at least in the Bayesian world; see Harvey for a
> frequentist treatment. Cites below.
>
> My essay on "Measurement" in the Oxford H'book of Political
> Methodology (Box-Steffensmeier, Brady and Collier, eds) surveys these
> models and others, and their application in political science settings.
>
> On the specific question of assessing dimensionality, I'm not aware of
> any cheap or "quick and dirty" method for doing this in the dynamic setting
> (a la looking at the eigenstructure of the indicators in a
> cross-sectional setting). Stimson and his partners might be among the few
> political scientists that have considered the dimensionality issue in a
> dynamic setting, afaik.
>
> -- Simon Jackman
>
>
> @Book{west:bk,
> author = {West, Mike and Jeff Harrison}, title = {Bayesian
> Forecasting and Dynamic Models},
> publisher = {Springer-Verlag}, year = 1997, address = {New
> York}
> }
>
>
> @Book{harvey:bk,
> author = {Harvey, Andrew~C.}, title = {Forecasting, structural
> time series models and the Kalman filter},
> publisher = cup, year = 1989, address = ny }
>
>
> On Jun 9, 2009, at 5:30 AM, eric magar wrote:
>
>
>> Have you read about IRT estimation? IRT models, developed for the
>> educational literature, are designed to infer some latent variable, in
>> one or more dimensions, from a series of related indicators. A discussion
>> can be found in Gelman and Hill's volume. Hope this is useful.
>>
>> Eric Magar
>> Departamento de Ciencia Política
>> ITAM, México DF
>> +52 55 5628 4079
>>
>>
>>
>>
>> On Mon, Jun 8, 2009 at 8:46 AM, Ebru Altinoglu <[log in to unmask]>
>> wrote:
>>
>>
>>> Greetings,
>>>
>>>
>>> Any suggestion on how to assess dimensionality of multiple
>>> indicators with TSCS or panel data? (I am trying to construct a
>>> dependent variable using multiple indicators, and believe that they
>>> form more than one dimension.) I assume that regular factor-analysis
>>> won't work with TSCS data. Yet could not find any method to use
>>> instead.
>>>
>>> Thank you.
>>>
>>>
>>> Ebru
>>>
>>>
>>>
>>> --
>>>
>>>
>>> Ebru Altinoglu
>>> PhD Candidate
>>> Department of Political Science
>>> University of Michigan, Ann Arbor
>>>
>>>
>>> **********************************************************
>>> Political Methodology E-Mail List
>>> Editors: Xun Pang <[log in to unmask]>
>>> Jon C. Rogowski <[log in to unmask]>
>>> **********************************************************
>>> Send messages to [log in to unmask]
>>> To join the list, cancel your subscription, or modify
>>> your subscription settings visit:
>>>
>>> http://polmeth.wustl.edu/polmeth.php
>>>
>>>
>>> **********************************************************
>>>
>>>
>>
>> **********************************************************
>> Political Methodology E-Mail List
>> Editors: Xun Pang <[log in to unmask]>
>> Jon C. Rogowski <[log in to unmask]>
>> **********************************************************
>> Send messages to [log in to unmask]
>> To join the list, cancel your subscription, or modify
>> your subscription settings visit:
>>
>> http://polmeth.wustl.edu/polmeth.php
>>
>>
>> **********************************************************
>>
>>
>
> Professor Simon Jackman,
> Jan-August 2009
> Visiting Professor,
> United States Studies Centre
> University of Sydney, NSW 2006
> Australia
> +61 2 9036 9208 (w)
> +61 401 620 725 (m)
>
>
> Depts of Political Science & (by courtesy) Statistics,
> Stanford University, Stanford, CA 94305-6044, USA.
> http://jackman.stanford.edu
>
>
>
>
>
>
>
>
>
>
> **********************************************************
> Political Methodology E-Mail List
> Editors: Xun Pang <[log in to unmask]>
> Jon C. Rogowski <[log in to unmask]>
> **********************************************************
> Send messages to [log in to unmask]
> To join the list, cancel your subscription, or modify
> your subscription settings visit:
>
> http://polmeth.wustl.edu/polmeth.php
>
>
> **********************************************************
>
>
**********************************************************
Political Methodology E-Mail List
Editors: Xun Pang <[log in to unmask]>
Jon C. Rogowski <[log in to unmask]>
**********************************************************
Send messages to [log in to unmask]
To join the list, cancel your subscription, or modify
your subscription settings visit:
http://polmeth.wustl.edu/polmeth.php
**********************************************************
|
