Help with correct statistical methodology

Katzu

New Member
We are conducting a study assessing how a reduction in disease severity based on objectives measures correspond to increase in quality of life and if this relationsship changes over time.

To do this we have are using two measures one for the objectives which have a score ranging from 0-72 and one for quality of life ranging from 0-30. We do the measures at baseline, after 3 months, after 12 months and then each subsequent year. We would like to look at how the relationsship is from 0-3 months, from 3-12 months and then from 12-24 months and so on. Additionally we want to see what has the largest impact on quality of life; a decrease or increase disease severity based on the objective measures.

My question is, is there a smart way to do so? Or should you calculate the difference between for the measures between two time point and then do correlation analysis and then compare the coeffecients between the time periods?

Would it make sense to do linear regression instead and compare r^2 between the time points?

Karabiner

TS Contributor
You can regress QoL on "disease severity", but you have take into account
that measurements are clustered within the same persons.
I'd therefore suppose that it is a case for multilevel modeling. With
this, you can also include "time" and the "time*severity" interaction
(representing a possible change in the relationship between QoL and
severity over time).

With kind regards

Karabiner