Report a level of certainty with any subscore category
The advice. “You need to report some level of certainty in a category with some measure of error.”
Why. This is what makes Report the least bad information rather than none defensible rather than merely pragmatic. The committee agreed that Michigan will continue reporting subscores from a test that does not really support them. The condition it attached is that the report must carry its own uncertainty.
A category shown without error looks like a fact. The same category shown with the certainty attached to it is a different communication — and given that classification accuracy sits in the 0.10–0.35 quadratic weighted kappa range at realistic dimensional separations (CT-PO vs SBAC Simulation Results Deck), the uncertainty is not a footnote. It is the finding.
The existing defect this speaks to. For some subscores the At/Near Standard band cannot be drawn at all because there is only one data point. That is currently invisible to the reader.
Status. Open. This is the recommendation the promised sample score report (Bring a sample score report to the next meeting) should be judged against.
Related. M-STEP Subscore Reporting