[R] Inserting weights in ltm package

[Thomas Lumley]

On Wed, Jul 27, 2011 at 2:58 PM, Erica Crome <erica.crome at mq.edu.au> wrote:
> Afternoon R help,
>
> I want to run Rasch/IRT analyses using the ltm package, however, I am
> using large scale survey data which requires weighting for accurate
> results. I attempted to create a weighted object to insert into the
> formulae of the ltm packages, however, the survey data only includes
> 30 replicate weights and a sampling weight. The svrepdesign requires
> additional information  such as jacknife weights in an external object
> (for scale, rscale), however, I do not have these. I typically just
> use the sampling weight when doing analyses in other platforms such as
> MPlus, however there does not appear to a way of referring to this
> directly in the ltm formulae.
>
> My question is: am I able to insert sampling weights directly into an
> ltm analysis, or is there a way of creating a weighted data object
> with the information available which I could then insert into the ltm
> analyses? I have been unable to find any similar help topics using the
> websearch. Thanks.
>

You can easily use svrepdesign() to create a weighted survey design
object using the replicate weights, but that won't help you. The
functions in the ltm package do not appear to accept any sort of
weights.  I don't know enough about those models to know if simple
weighting would be enough or whether pairwise sampling probabilities
would be needed as they are for linear mixed models.

   -thomas

-- 
Thomas Lumley
Professor of Biostatistics
University of Auckland