3

I have responses to a questionnaire item from a number of people, measured at equidistant timepoints. I wish to fit a growth mixture model (in R, using the LCMM package) to this data to find latent classes. My data looks something like this:

    ID  item-response  timepoint
    -----------------------
    1       3           1
    1       2           2
    1       2           3
    2       2           1
    2       3           2
    2       2           3
    2       1           4
    2       1           5
    2       3           6
    2       2           7
    2       2           8
    2       2           9
    2       1           10
    2       4           11
    2       2           12
    3       1           1
    3       1           2
    3       1           3
    3       1           4
    3       1           5
    .       .           .
    .       .           .
    .       .           .

The item is one of 13 on a questionnaire on mood states. Responses are given on a Likert-scale (1 to 5).

A plot of the response curves of the first four individuals looks like this: Response curves of the first four individuals

I am worried about the fact that the number of measurements per person is not the same. Is this a huge problem for growth mixture models or not so much?

[edit] included a column of timepoints

Stijn
  • 1,550
  • 1
  • 12
  • 20
  • What is the x-axis in this plot? Is there a "time" associated with each of the repeated measurements? – Macro Jul 11 '13 at 15:33
  • The x-axis denote time. Measurements were taken each week. – Stijn Jul 11 '13 at 16:09
  • The layout of your data set at the top of the post does not include the time variable. Do you have this information? – Macro Jul 11 '13 at 17:09
  • I've added a column of timepoints to make the data more clear. As you can see, the person with ID 1 only has three measurements, while ID 2 has twelve. So at some point, certain individuals stopped, while others kept providing measurements. (Question: do we call data like this *censored data*?) – Stijn Jul 11 '13 at 19:03
  • 1
    Are these data responses to a Likert-type item? If so, could you please tell us a little more on the questionnaire that's being used? – chl Jul 11 '13 at 19:55
  • Ah, sorry. I forgot to mention it's Likert-type data. I've updated the question. – Stijn Jul 11 '13 at 23:16
  • I would be mainly concerned by measurement error at the level of a single item, not with unbalanced data unless you can't assume data are missing at random. – chl Jul 12 '13 at 10:24

0 Answers0