I was asked about if/when I would use the median vs. most recent information in determining a grade. Statistically speaking, the median is a better measure of central tendency for grades (vs. using an average)---but so often with standards-based grading, we just talk about examples where most recent information is best.
Here is where I use the median and not the most recent information: content standards.
For example, here is one of the Washington State Grade Level Evidences (GLEs) I am teaching to this semester:
1.2.6 Understand cellular structures, their functions, and how specific genes regulate these functions.
- Describe cellular structures that allow cells to extract and use energy from food, eliminate wastes, and respond to the environment (e.g., every cell is covered by a membrane that controls what goes into and out of the cell).
- Describe how DNA molecules are long chains linking four kinds of smaller molecules, whose sequence encodes genetic information.
- Describe how genes (DNA segments) provide instructions for assembling protein molecules in cells.
- Describe how proteins control life functions (e.g., the proteins myosin and actin interact to cause muscular contraction; the protein hemoglobin carries oxygen in some organisms).
What's in bold is the target. The bullets underneath are examples, but are not a "to do" list. There is plenty more that could be there.
In this case, it makes no sense for me to look at improvement on this target. It's too freaking big. I have to teach to this standard in chunks. Even just looking at the bullets, we've got cells, DNA, protein synthesis, and protein function. It's going to take me about half of the semester to help kids navigate this...and at the end, I need a gestalt. I need to know if each kid is at the boldfaced standard. This is where the median will come in. I'll have a variety of summative assessments for this GLE collected throughout the semester. I don't need to know specifically how kids did at the end (if they showed improvement), because each chunk is different.
Skills, however, are perfect for use with trend data. One would expect a student to improve over time---whether it's writing a hypothesis, speaking a foreign language, or changing the oil in a car. The first few times may not be so good, but with practice, the kids should be hitting the target (and teachers can throw out those early attempts).
The $100K question is how to "crunch" everything into a single grade...something I'll comment upon tomorrow.