Hi Statlisters,
I am currently using Stata 18 version.
I am fitting a mixed effects model with splines to a longitudinal biomarker. Because it is health related, more measurements are done when the health of the patient is worsening. Therefore I tried to implement a method that removes a random sample of points from 'clusters'.
I.e. clusters of points are defined as a lot of measurements in a very short time, and I keep for example 50% of those points. This is to adjust for the fact that worse health causes more measurements to be taken, and this may bias the modelling of the variable in the direction of the more sick patients.
My question is: How do I now verify if this approach works/is better? Because I am now using a different dataset for both models, as explained above (one have less values), likelihood tests are not possible, then which options I have?
Thanks in advance!
I am currently using Stata 18 version.
I am fitting a mixed effects model with splines to a longitudinal biomarker. Because it is health related, more measurements are done when the health of the patient is worsening. Therefore I tried to implement a method that removes a random sample of points from 'clusters'.
I.e. clusters of points are defined as a lot of measurements in a very short time, and I keep for example 50% of those points. This is to adjust for the fact that worse health causes more measurements to be taken, and this may bias the modelling of the variable in the direction of the more sick patients.
My question is: How do I now verify if this approach works/is better? Because I am now using a different dataset for both models, as explained above (one have less values), likelihood tests are not possible, then which options I have?
Thanks in advance!
