When the Mean is Misleading: a Guide to Ordered Regression for Meaningfully Modeling Ordinal Outcomes