Maarten, you are right. I tried to estimate using the unimputed
dataset with the command:
xtmixed dvar if [parwgt] || schid: || childid:,
and it has the same problem (no standard errors for random
components). The model I am estimating has three levels: time, child
(using the variable "childid"), and school (using "schid"). However,
children are not entirely nested within schools, because children
switch schools over time. So I tried the crossed random effects models
using:
xtmixed irtm if [parwgt] || _all: R.schid || childid:
But it gives error message like this:
J(): 3900 unable to allocate real <tmp>[38153,2413]
_xtm_mixed_ll_uu(): - function returned error
_xtm_mixed_ll_u(): - function returned error
_xtm_em_iter_u(): - function returned error
_xtm_em_iter(): - function returned error
<istmt>: - function returned error
r(3900);
Now I am not sure how to specify the random effects at the school
level since neither nested model nor crossed model seems to work. Any
suggestions?
Thanks!
On Thu, Apr 23, 2009 at 10:26 AM, Maarten buis <[email protected]> wrote:
>
> --- Nailing Xia wrote:
>> Thanks, Maarten. I did what you suggested, and it happens that when
>> estimating the third imputed dataset, it keeps doing the iterations -
>> showing "Iteration xx: log restricted-likelihood = -xx.xx (backed up)"
>> again and again. I guess this is the sign for the model being
>> unidentified in the third dataset?
>
> Yes
>
>> However, even if I drop the third imputed dataset, there is another
>> problem with other imputed datasets. The result only includes
>> coefficient and standard error for the intercept, and coefficients for
>> the random components, but no standard errors for the random effects
>> parameters. It also shows a warning, saying that "convergence not
>> achieved; estimates are based on iterated EM". Is this also an
>> identification problem? What can I do about it?
>
> That just looks like there is something very wrong with your -xtmixed-
> model. I would focuss of fixing that before doing a multiple imputation
> model. My first suspect would be the two levels of nesting: are they
> correct, or did something go wrong while preparing the data? Do you
> need both levels? Once you got -xtmixed- to work with the unimputed
> data, try it again on the imputed data. If it still doesn't work, check
> the imputation model. Again my prime suspect would be the nesting: did
> the model properly acount for the nested structure? Did it do something
> weird to the level identifiers?
>
> -- Maarten
>
> -----------------------------------------
> Maarten L. Buis
> Institut fuer Soziologie
> Universitaet Tuebingen
> Wilhelmstrasse 36
> 72074 Tuebingen
> Germany
>
> http://home.fsw.vu.nl/m.buis/
> -----------------------------------------
>
>
>
>
> *
> * 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/
>
*
* 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/
