Am going to analyze population based mental health study that has used GHQ 12 (Likert scoring 0-1-2-3). Want to categorize GHQ 12 into three categories for ordinal logistic regression.
It sounds like your measure will produce a continuous score, so why would you want to "lose information" by reducing it to cut-off scores -- especially when a simple linear regression on that dependent variable would be much easier to interpret than an ordinal logistic regression.