06-15-2017 01:17 PM
I have a question regarding a longitudinal analysis for measurements over time. Let's say I'm interested in assessing the effect of age at study entry on blood pressure. Participants will have different numbers of measurements and measurements will not necessarily be taken at the same intervals (ie one patient may have measurements taken at 1 month, 9 months and 17 months, another patient may have measurements taken at 1 month, 2 months, 3 months, 4 months, etc). Time is calculated as date of measurement minus study entry date. I'm doing this analysis in proc mixed (SAS 9.4) and using a spatial power structure to model the correlation within patients (ie as measurements get farther apart they become less correlated).
My question is: Does it make sense to look at a univariate analysis where I have only adjusted for age at study entry, or should this type of analysis always include time as a covariate to be properly interpreted?
Thank you for taking the time to consider my question!
06-15-2017 04:31 PM
I'd say, it depends.
Do you have a research question about how blood pressure changes over time? Do you expect that mean blood pressure will change over time? Are different treatments applied to patients, and if so, might the time profile be affected by treatment?
Or are the multiple measurements merely subsamples?