One Teaching Method Pulls Ahead, and Two Are Tied

Flipped instruction beat both alternatives by about eight exam points; Lecture and Blended performed the same.

To Department Chair

From Teaching & Learning Analytics

Re Three-method pilot, final exam scores

Where this comes from

Chapter Chapter 164 · One-Way ANOVA: Three Teaching Methods

Part Part XXVII · Capstone Projects: Statistical Testing

Dataset capstone-teaching-methods-exam-scores.xlsx

Notebook View the analysis

Recommendation

Bottom line

Expand the Flipped format, and treat Lecture and Blended as interchangeable on exam performance. Flipped students scored about 8 points higher on the final than students in either other format, a gap far too consistent to be chance. Before committing, confirm that students did not choose their own section, because self-selection would produce this same result without the format doing any of the work.

What we found

We compared final exam scores for 130 students across three formats. Lecture averaged 73.2, Blended 73.5, and Flipped 81.1. The overall test says the three are not alike (p is well under 0.001), and the follow-up comparison locates the difference precisely: Flipped beats Lecture by about 7.6 points and Blended by about 7.9 points, while Lecture and Blended sit within half a point of each other.

Two panels: group means with confidence intervals showing Flipped clearly higher, and Tukey pairwise differences where the two Flipped comparisons clear zero while Lecture versus Blended straddles it. — Figure 1. Left: average score per method with its 95% range. Flipped sits clearly above the other two, which overlap almost entirely. Right: each pairwise gap with its range; the two involving Flipped clear the zero line, the third does not.

Bar chart: lecture 73.2, blended 73.5, flipped 81.1. — Figure 2. Mean exam score by format, on a truncated axis so the gaps are visible.

Why the two-part answer matters

It would be easy to report only that 'teaching method made a significant difference'. That sentence is true and misleading. It invites the reader to assume all three formats differ, when in fact two of them performed identically. The actionable finding is narrower and more useful: one format helps, and choosing between the other two on exam grounds alone is a coin flip.

How much of the story is method?

Teaching format explains roughly 15% of the variation in exam scores. That is a substantial share for an educational intervention, but it also means the large majority of the difference between students comes from everything else, prior preparation, effort, and circumstance. Adopting Flipped should be expected to shift the average, not to level the class.

Before you act on this

Two cautions. First, and most important, this only supports a claim about the format if students were assigned to sections rather than choosing them; if the Flipped sections attracted more motivated students, we are measuring the students. Second, instructor and format are entangled here, so if one instructor taught all the Flipped sections we cannot separate the two. A follow-up that rotates instructors across formats would settle both questions.

← Back to Chapter 164 📓 Companion notebook 📊 Download dataset

From Statistics, Data Science and AI: A Visual Handbook by John Fisher. Every statistic, table, and figure in this report is reproduced by the companion notebook.