use continuous data or group based on score?

Question

I have done a research where I grouped participants based on their scores on Beck Depression Inventory (BDI-II). In study 1, my groups were healthy controls (scores 0-13) vs depressed (>13). In study 2, they were regrouped to mild, moderate and severe depression in addition to the control group. I've been under the impression that what I do is fine, but reviewers in a journal I send the paper for have mentioned that it reduces the power of study and that it is "unconventional" to group based on BDI scores (and not clinical interview). They have suggested that I erase the groups and re-do the analysis with the continuous BDI-II scores. But doing this would mean completely rewriting the paper.

I need opinions from people who are better than me in statistics. Is it worth re-doing the analysis? Does it reduce the power really that much? is it really that unconventional? I will be very grateful to get your opinions. thank you in advance!

Answer 1

Grouping your observed data reduces power because it erases all differences within the groups.

For example, if the main effect of the intervention is to reduce depression from 12 to 4, but will not affect depression at levels 12 and above, grouping the data would erase that kind of difference.