Specifications Grading for Success in Statistics Education
Wesley S. Burr
Trent University
Why Do We ‘Grade’ Students?
I have four questions for you to consider.
- What is the purpose of assessing students’ work?
- Where did grades come from?
- How do we grade, and cui bono?
- How do we resolve the conflict between grade stakeholders?
An overly hyperbolic statement
I do not know one (statistics) colleague who genuinely enjoys the act of grading undergraduate student work. It is drudgery at best; hellish at worst.
What is Specifications Grading?
- book released in 2014
- I became aware of it via a colleague (K. Kinnaird, at ICOTS10 in Kyoto!) in 2018
- tried it the following semester
- immediate convert to the concept
- (courtesy of dinner with Helen) a modern version of the Keller Plan (1965-68)
How Far Has This Spread?
- numerous converts in the mathematical, statistical and computational sciences
- some publicity in 2016, then a surge of interest after 2020
- some links for those interested
My Experience
I teach a broad cross-section of 2nd through 4th year undergraduate courses, and occasional graduate specialist courses. I have implemented variations of specifications grading in:
- 5 iterations of Mathematical Statistics (2nd year)
- 2 iterations of Stochastic Processes (4th year)
- 3 iterations of Linear Models (3rd year)
- 2 iterations of Experimental Design (4th year)
- 1 iteration of Statistical Learning (3rd year)
All from 2018 Fall through 2023 Winter - 5 years.
How does it work: Mathematical Statistics
- broke course down into the most atomic learning outcomes feasible
- in its latest iteration:
- a total of 30 target concepts, with 15 accompanying challenge areas (45 modular pieces)
- students work independently, submitting work weekly
- any item can be redone: all are graded on successful/unsuccessful (S/U)
- maximum number of submissions per week
Mathematical Statistics: Example Topics
- Target Concept 13: Sampling Distributions I
- Target Concept 21: Bootstrap-t CIs for Means
- Challenge Problem 9: Parametric CI Simulation
How does it Work: Linear Models
second major course, Linear Models, is a 3rd-year course on linear models for statistics majors
students largely choose to be in the class,
more mathematically and academically mature than the students in Mathematical Statistics
the course is broken into modules, each of which is fairly broad
each module has a bank of problems
students complete a set number of problems from each module (e.g., best 6 count)
they may do as few or as many problems as they wish
each problem is graded on a NA/Bronze/Silver/Gold scale
problems may not be redone, but additional problems from the module may
Advantages to Both
- grading is very simple - specifications make for extremely easy and rapid assessment of student performance
- students have full control over their final grades, based entirely on their performance on the assessments
- significantly reduced grade and test anxiety
- feedback feels meaningful but not punitive
- most students really like the system
Disadvantages to Both
- both systems are more complicated to implement
- preparing larger banks of problems, and refreshing them, is higher workload
- students can fall behind, and have no effective way of ‘catching up’
- depends heavily on student engagement and academic integrity
Is It Worth It?
I can’t imagine going back. Students are happier, I’m happier, and the students’ overall comprehension is either higher, or at least comparable at worst.
I don’t hate grading when it’s done in this framework.
If you are interested in trying a variation, reach out: I’d be happy to share my materials and details of my implementations.
