External Email - Use Caution

Hello FreeSurfer Developers, I have a general statistical question regarding regression models.

Our team attempts to create predictive models based on FreeSurfer outputs such as volumes, thicknesses, and curvatures. The predictive variables we are using include several demographic and metabolic factors. We have adjusted our models to account for intracranial volume (ICV) in the population. Instead of simply dividing our outputs by the ICV, we found through our literature review that it is a better approach to include ICV as a covariate alongside our predictive variables in the model.

However, we are concerned that including ICV in this manner leads to a significant increase in the R-squared (R²) value of our models due to the effect size provided by ICV. I would like to know whether this is a statistically significant issue, and if it is, how we should address it. Thanks again for your valuable help, Yunus