Announcement

Collapse
No announcement yet.
X
  • Filter
  • Time
  • Show
Clear All
new posts

  • linear mixed model with more than two dependent variables

    Hi.
    I tried to search an answer in previous posts but i didn't find anything.. anyway i apologize if i'd sound repetitive!

    I have a database with 30 observations (patients), each of them evaluated with various (about 10) clinical evaluation scale, repeated at three time points (before, after treatment and after a period of follow up - first and second time point are fully filled, at third there are some missing values).
    I divided patients into three different groups and i want to test if there are differences within and between three groups at second and third time point.

    Searching into stata manual and on statalist i found many information about using xtmixed with one dependent variables and interactions between independent ones and time-points but..
    if i have more than one dependent (in my database, multiple scores of evaluation scale - continuous and not -) i must run a mixed model for each one or is there a command to perform a single analysis?

    Thank you

    Emanuele

  • #2
    Emanuele:
    welcome to the list.
    As far as i now, you should perform different regression model for each dependent variable.
    But I would take a step aside, first: due to your limited sample size (30 obs), I would not expect to obtain that much out of your data, regardless the regression approach you have in mind.
    Kind regards,
    Carlo
    (Stata 18.0 SE)

    Comment


    • #3
      Hello Emanuele,

      Carlo gave excellent advice.

      I just wish to add that, since your sample size is quite small (as Carlo pointed out); what is more, you wish to divide the sample into 3 groups; moreover, you have missing data related to the main endpoint (the last phase of the repeated measures); last not least, you wish to perform multiple comparisons, hence incurring (potentially) in familywise error, I wonder whether you shouldn't think about "putting to a test" a Bayesian model.
      Best regards,

      Marcos

      Comment


      • #4
        Thank you for your advices.

        So i have more than an issue in my analysis. Carlo, how would you treat my data?
        Maybe i could limit to analyze only first two time points? Excluding the third, linear mixed model would be still adequate ?

        Comment


        • #5
          Emanuele:
          Marcos pointed out some other interesting issues concerning your dataset.
          With such a limited sample size, provided that cannot be increased, there's probably little room for any inference.
          Anyway, you may want to take a look at -sureg-.
          Kind regards,
          Carlo
          (Stata 18.0 SE)

          Comment

          Working...
          X