Modelling change for an uneven interval between first and second measurements

[Sophie Hutchinson]

Hello Stata Forum

I have a series of baseline measurements, which I am interested in using to model the change in an outcome variable.

The outcome variable is measured twice: once at baseline and once at an irregular interval after that (around two to four years per person). The predictor variables are measured only at baseline.

I am currently modelling the change per year as the outcome variable in a linear regression.

Would it be better to model absolute change and have years between measurements as an additional predictor variable? Or to take an ANCOVA approach and model the second measurement from the first measurement plus the years between measurements (and other predictor variables)? Or some other approach I have not thought of?

Kind regards

Sophie

[Phil Bromiley]

You didn't get a quick answer. You'll increase your chances of a helpful answer by following the faq on asking questions - provide Stata code in code delimiters, readable Stata output, and sample data using dataex.

How you measure your dv (change, absolute change, etc.) is really about the substance of your research - there is no general solution irrespective of the details of the analysis. For example, it would make a difference if you have something you think changes once or you have something where the x's result in continuing changes each year. If it is the later, then change per year makes sense - controlling for years does not take care of the issue that the parameter on other iv's should differ given different number of years observed.