Statalist


[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index]

Re: st: RE: RE: RE: Having multiple data files open in Stata 9


From   Austin Nichols <austinnichols@gmail.com>
To   statalist@hsphsun2.harvard.edu
Subject   Re: st: RE: RE: RE: Having multiple data files open in Stata 9
Date   Fri, 25 Sep 2009 14:04:47 -0400

Barth Riley--

Under Nick's option (1) of "not aligned" data falls my approach to
matching across datasets using an unmatched -merge-
[see e.g. http://www.stata.com/statalist/archive/2009-08/msg01521.html
], which is equivalent to having two datsets in memory at once, as
long as no variable names are shared across the datasets (once you
have both "datasets" in memory side by side like that, you can also
access each "dataset" in Mata if you prefer).  But if you want to
simulate item responses based on IRT item parameters stored in one
dataset and person parameters stored in a second dataset, you probably
need to -merge- or -cross- so that each observation is an item-person
pair, i.e. each observation records this item's characteristics and
this person's characteristics.  If you have 3 thousand people and 1
thousand test items, this will result in 3 million observations--but
do you really want this many simulated observations? If your goal to
produce some summary measure by person or item, there is probably
another way to do it.

If you are *estimating* an IRT model, be forewarned:
http://www.stata.com/statalist/archive/2009-03/msg00599.html

On Fri, Sep 25, 2009 at 1:17 PM, Nick Cox <n.j.cox@durham.ac.uk> wrote:
> As others have pointed out, that doesn't follow at all. In addition to
> other points,
>
> 1. It is perfectly possible to have some variables in the dataset that
> are not aligned with other variables. Stata doesn't know what variables
> mean and doesn't care. So, for example, people often have time series
> for times (... 1952, 1953, ..., 2008, whatever) and also work with lags
> 1, 2, 3, whatever. Variables for times are not aligned with variables
> for lags. It's true that (a) you need to work out how to get the
> different subsets to talk to each other (b) this may not be very
> efficient (c) you need to take care not to produce nonsense.
>
> 2. Mata.
>
> 3. In similar problems I tend to have (my analogue of) the person data
> in memory and use something like -postfile- or -file- to store
> simulation results.
>
> Nick
> n.j.cox@durham.ac.uk
>
> Barth Riley
>
> Thanks for your replyMartin. Between the fact that Stata only allows one
> datafile open at a time and the fact that it doesn't support arrays
> makes
> certain tasks very difficult to do. I will most likely have to do part
> of
> the work in another application.
>
> Martin Weiss
>
> It is pretty much the same issue for all Stata versions. There is only
> one
> dataset open at any time, even though you can of course -merge- and
> -append-
> as much as you like. You can also have multiple instances open, but they
> would not be able to communicate with each other...
>
>
> Barth Riley
>
> Is there a way to have multiple datasets open at one time in Stata
> version
> 9? I want to simulate item responses based on IRT item parameters stored
> in
> one dataset and person parameters stored in a second dataset.
>
*
*   For searches and help try:
*   http://www.stata.com/help.cgi?search
*   http://www.stata.com/support/statalist/faq
*   http://www.ats.ucla.edu/stat/stata/



© Copyright 1996–2014 StataCorp LP   |   Terms of use   |   Privacy   |   Contact us   |   What's new   |   Site index