Re: st: can GLLAMM handle missing data?

[David Hoaglin <dchoaglin@gmail.com>]

Hi, Francesca.

-gllamm- and other commands for longitudinal data do use all the
available data, not just the "complete cases."  Linear mixed-effects
models and generalized linear mixed-effects models have the advantage
of being flexible and accepting data in which some individuals do not
have data at all time points.

Whenever data are missing, however, the analyst must investigate the
reasons --- the process(es) that caused the missingness.  Often, using
complete cases or the available data will produce biased results.
Missing data and approaches for handling them have an extensive
literature.  One approach, multiple imputation, is nicely implemented
in -mi- in Stata.

A number of books discuss analyses of longitudinal data and issues
surrounding missing data in that setting.  The book by Fitzmaurice,
Laird, and Ware (2011) is accessible and fairly comprehensive.

David Hoaglin

Fitzmaurice GM, Laird NM, Ware JH (2011).  Applied Longitudinal
Analysis, 2nd ed.  Wiley.

On Wed, Mar 26, 2014 at 7:34 AM, Pesola, Francesca
<francesca.pesola@kcl.ac.uk> wrote:
> Hi Nick,
>
> Thanks for your reply.  Apologies for such a simple question but I am new to GLLAMM and I was just surprised that the output states:
>
> number of level 1 units = 2284
> number of level 2 units = 566
>
> these figures reflect the sample size for those respondents who have data on at least 1 data point, which is why I assumed it included them all rather than just those with complete cases on all time points.
>
> If it does remove missing data, would it do it listwise? How can I find out how many cases are included in the analysis?
>
> Thanks,
> Francesca
*
*   For searches and help try:
*   http://www.stata.com/help.cgi?search
*   http://www.stata.com/support/faqs/resources/statalist-faq/
*   http://www.ats.ucla.edu/stat/stata/