# Three-tier rubric skeleton

Version 1.0, August 2026. Adapted from the final report rubric I use in EPI 553 (Principles of Statistical Inference II).

## Why tiers rather than adjectives

A rubric that says "thorough," "adequate," and "insufficient" is not a rubric. It is a vocabulary for defending a grade you already assigned. Two graders using it will disagree, and the student cannot use it to check their own work before submitting, which is the only thing a rubric is actually for.

The fix is to write, for each component, the observable difference between the three tiers. That takes longer once and is faster every time after, because the grading argument has already happened, on paper, with yourself.

## The pattern

| Component (pts) | Full credit | Partial credit | No credit | Pts |
|---|---|---|---|---|
| **[Name]** (n pts) | [Every required element, listed. This column is a checklist, not a compliment.] | [The realistic near-miss. Name the specific one or two failures you actually see, joined by "or".] | [The floor. Usually absent, or present but unusable for the stated purpose.] | n |

Three things make this work.

**The full-credit column is a checklist.** If you cannot write it as a list of things that are present, the component is not well defined yet.

**The partial-credit column names the near-miss you actually see.** Not "some elements missing" but "one variable's recoding not described, or missing data percentage not reported, or software version omitted." Students recognize themselves in that sentence, which is the point.

**The no-credit column separates absent from unusable.** "Methods section is absent or too vague to reconstruct the analysis" tells a student that a present-but-vague methods section earns nothing, which is a thing they do not otherwise know.

## Worked example

From the EPI 553 final report rubric, the Methods row, 7 points:

- **Full credit.** Data source and design described; exclusion criteria documented with counts; all variables described with scale and recoding; missing data quantified with strategy stated; statistical analysis plan is complete and reproducible.
- **Partial credit.** Most elements present; minor gaps, for example one variable's recoding not described, or missing data percentage not reported, or software version omitted.
- **No credit.** Methods section is absent or too vague to reconstruct the analysis.

And the Discussion row, 5 points, which is where the adjectives usually creep back in:

- **Full credit.** Main finding restated; interpreted in context of prior literature; confounding and bias discussed; specific limitations named with explanation of impact; implications stated.
- **Partial credit.** Discussion present but superficial, for example generic limitations, no literature comparison, or implications absent.
- **No credit.** No discussion section.

Note what "generic limitations" is doing there. In the assignment description that pairs with this rubric, the corresponding line reads: generic statements such as "this study has limitations" do not earn credit; name the specific limitations. The rubric and the instructions have to agree, in the same words, or students follow the one that is easier.

## Building it

Write the full-credit column for every component first, in one pass. Then write no-credit for every component. Partial credit last, because by then you know what the gap between the two actually contains.

Budget about ninety minutes for a rubric covering a major assignment with eight to ten components. It carries forward to every subsequent offering with minor edits.