I am planning to run separate regression analysis with cortical thickness in two very different populations (young and elderly subjects). Depending on results, it may happen that I am interested in comparing regression slopes between the two groups. My question is: should I build one average subject for each within-group regression analysis (one for young and another for elderly), and another different average subject merging all subjects of the two samples to compare regression slopes? Or I should preferably use only one average subjects with all subjects in all within and between regression analysis?

Many thanks in in advance.