Hi Statalisters,
I wanted to ask regarding growth mixture modelling for register data (PS I am not statistician
).
Observations are recorded at different measurement points and as some patients have multiple entry the
number of "waves" are not the same across observations. As the questionnaire consist of followup (probing)questions, you can imagine that there will be different measurement points and different “waves” for each client (with different age), and that there will be a number of missing values across variables (questions). I have tried to illustrate this with the following table below my questions. My questions are:
Regards
I wanted to ask regarding growth mixture modelling for register data (PS I am not statistician
).Observations are recorded at different measurement points and as some patients have multiple entry the
number of "waves" are not the same across observations. As the questionnaire consist of followup (probing)questions, you can imagine that there will be different measurement points and different “waves” for each client (with different age), and that there will be a number of missing values across variables (questions). I have tried to illustrate this with the following table below my questions. My questions are:
- does the concept of treating time more flexibly apply for such data with big measurement point difference?
- Is there appropriate (latent} growth curve model which I can specify by using Stata, for example if I want to measure changes of composite scores (continuous variable) across time or practitioner severity rating (Likert Scale type 0-9) across time?
- If I want to have trajectory (spaghetti plots), what will be the X axis (as measurement points vary across observations)?
- Is it possible to work with missing values (excluding imputation)across variables without dropping observations with missing values while doing analysis, instead of complete case analysis ?
- Is there any other Stata procedure which might be relevant to analysis of the health outcome trajectories of individuals and latent classes (identified from baseline data) across, lets say, 10 years (here not all observations have same interview dates nor they have same number of repeated measurements)
| id | Age | Interview intake order | Interview date |
| 01 | 50 | 1 | Dec 2009 |
| 01 | 52 | 2 | June 2012 |
| 01 | 54 | 3 | Aug 2014 |
| 02 | 30 | 1 | Apr 2008 |
| 02 | 32 | 2 | Aug 2010 |
| 03 | 20 | 1 | Dec 2003 |
| 03 | 21 | 2 | Mar 2005 |
| 03 | 21 | 3 | Oct 2005 |
| 03 | 28 | 4 | Sept2012 |
