Re: st: test of repeated measurements - ordinal data (likert scale)

[David Hoaglin <dchoaglin@gmail.com>]

Hi, Mette.

The patients' responses to that question produce an ordered categorical outcome.

You should be able to use -ologit- to take into account the clustered
nature of the data (i.e., repeated measurements).

I suggest that you not focus too early on tests.  Start by making
various plots of the data, to discover what form the trends take.
Rabe-Hesketh and Skrondal (2012, Chapter 11) develop an example that
shows the process, including the use of -ologit- with clustered data.
(It's also in Chapter 7 of their second edition.)

David Hoaglin


Rabe-Hesketh, S. and A. Skrondal (2012). Multilevel and Longitudinal
Modeling Using Stata, third edition.  College Station, TX: Stata
Press.

On Fri, Jul 6, 2012 at 4:14 AM, Mette Olsen <meo@life.ku.dk> wrote:
> Hi,
>
> I wish to find a test that can compare patients' evaluation of a nutritional supplement over time.
> Patients are asked the same questions (e.g.. "how did you overall find the supplement?" - 'dislike very much' 'dislike' ' indifferent' 'like' or 'like very much') after 2 weeks, 1, 2 and 3 months of supplementation, respectively.
> I wish to know if there is a trend towards improved/worsened evaluation during the supplementation period, but I am not sure which type of test I should use. I have been suggested using xtmixed models but am not sure if the test allows for non-parametric data? I've also been suggested ologit but I'm not sure it can be used for repeated measurements in the same group.
>
> I look forward to learning more! Thanks in advance.
> Mette Frahm Olsen,
> PhD stud

*
*   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/