Monday, May 25, 2015

Even Retake Quizzes Predict Test Success

If I had to summarize this academic year in one word, it would be mastery.  Following the lead of my PLC teammate, and with the support of our department's intern teacher, I learned a ton about mastery learning and assessment.  It would take a few dozen blog entries to tell the whole story, but I'll start with something interesting: the strong correlation between cumulative quiz results and test results in the mastery setting.  These results held true in both a 5th grade classroom where I provided some assessment support and my high school Geometry class.

Generally speaking, quizzes SHOULD predict test scores if they are in proper alignment.  However, in our mastery classrooms, students have a lot more control over their quiz scores.  First, they are self-graded.  There are not a lot of opportunities to be dishonest the way the classroom is setup, but I am continually impressed at how accurately most students self-check their work.  Second, students can take unlimited retakes.  This means that they can nearly guarantee a score of 4/4 (meaning no mistakes) if they are willing to keep working and seek help to make improvements between quizzes.

5th grade: multiple choice mastery quizzes (by standard) predict MCA (state NCLB assessment) test scores, r=0.91.  Reaching 60% proficient in the Mastery Connect quizzes that we created nearly guaranteed MCA proficiency with only a few exceptions in the 60-70% range.  Every student below 55% on the Mastery Connect quizzes was not proficient on the MCA.  This means that there is no longer guesswork needed for interventions -- students can be reliably identified by January as unlikely to pass the state assessment and can receive necessary remediation.

Geometry: quiz totals by unit (mastery) predict test scores.  Over 4 units, correlation ranges from r=0.66 to r=0.81, with most on the higher end (see graphs below).  When talking about proficiency, I made the cut-line of an average quiz score of half 3's and half 4's (3.5 average).  See the table immediately below for number of students who were proficient on the test (above 80%) when they were above or below the cut-line.  In the earlier units, this did not prove to be as predictive or critical.  By the last unit, the quiz scores essentially told you if you would be proficient or not.

Below 3.5 quiz avgAbove 3.5 quiz avg
Alg Expressions7/197/7
Alg Equations4/177/9


I like mastery-based quizzing because it gives students a sense of self-control over the course outcomes.  If they want a high quiz grade, they can adjust their effort level and actually achieve it.  More importantly, the high correlation with tests means that, by proxy, effort is highly correlated with test scores and thus overall course grade.  Students have the ability to control their outcomes more than ever.  Next step: communicate these results to students and continue to help them overcome the fixed mindset that leads many students to delay their first quiz or not even attempt re-quizzes.  Second step: rethink the context that motivates students to WANT to succeed, even if it takes extra effort, in order to encourage students to work harder than they ever have.