I want to start a study that repeatedly measures a factor across time and looks at its change by certain patient subgroups. Of course I want to account for dependence between observations at different times within each patient. However, the study is open cohort and so I do not know if it meets the definition of a longitudinal study as this seems to involve studying the same individuals over time. Are Generalised estimating equations (GEE) better in this scenario since it gives a marginal instead of an individual-level result?
In addition, I am using medical record data so the time points can only be defined at one year intervals but patients may have measurements on the factor at any time during the year.
Spencer - if you original intent is to answer an overarching question that requires more that one methodology, method, design, triangulation etc - then it can be considered mixed methods. If not - it's not mixed methods. The attached chapter may assist,
Your should consider Cox regression with mixed effects. Whenever chance of occurring/observing an outcome changes over time, you should account for time dependency in statistical analysis, and both methods (GEE & mixed methods) fail to account for.
More on this: http://onlinelibrary.wiley.com/doi/10.1111/insr.12214/full
Thank you for your answers Sandro, Diego, and Dean. If I can back up for just a second. I would like to look at repeated measures of a score and its change of over time.The value of the score is of interest and could be expressed as a mean or slope. Then, I would like to look at the same change in different subgroups of patients. All together, it is a descriptive analysis. I would love to stset (STATA) and begin follow-up time at the date of diagnosis for all patients but can I get a mean change or slope change using Cox regression or something similar?
r u trying to do a pretest and post test analysis type of study?
Spencer, it seems to me that your dependent data is a continuous/scale variable, unrelated to time of observation. If this is so, you could go with MANOVA.
https://www.stata.com/manuals13/mvmanova.pdf
- Badges
- Science method
- Similar topics
- Biological Science
- Ecology