|
June 15, 2002
The Administrator's Guide to Data-Driven Decision Making (cont'd)
STEP 3 Analyze the data
You don't have to wait until you've collected all your data before you start analyzing it. Even if you don't have all of the information you'd like up front, you can still observe important trends that will lead to more informed decisions once all the information is available. The practical side of data analysis involves helping teachers and administrators learn how to interpret data and respond with the best resources and strategies for implementation.
Technology training
Teachers and administrators need to be trained in two areas to give them the necessary skills to become good decision makers: they need to know fundamental spreadsheet and database techniques such as filtering, sorting, and creating pivot tables and histograms; and they need to be comfortable with fundamental data analysis concepts such as correlation and causation.
Asking the right questions
Mastering decision-support technology is secondary to the critical skill of asking good questions of your data. Data-driven decision making is all about correlating data elements and exploring those factors that contribute both positively and negatively to student and teacher performance. For example, an incisive data-mining question might investigate the relationship between the number of library books a student checks out and their SAT-9 score in eighth grade. If there is a relationship, determine if the relationship is causative or correlative. In other words, if students who check out more books tend to score higher on the SAT-9 than those who check out fewer books, does that mean that reading more library books helps students score better? Finding causative relationships can lead to specific remedies.
Once a relationship is found, follow up with a series of questions that explore all aspects of the relationship, looking for gaps and inefficiencies. If reading more library books helps improve test scores, then is performance also related to the type of books students read? Does student performance rise or fall if they're forced to check out books during a weekly library period versus checking out books by their own free will? Questions such as these may lead to investigations into the quality of the library collection or the amount of time in the schedule for free reading.
While drilling down into data can be rewarding, often the further you drill down into a set of data, the smaller the sample size available, and as a result, the less accurate your conclusions. While your drill-down into the data may show that students who read 25 or more fiction books per year outperform students who read 25 or more nonfiction books, there might be only two or three students in each group of readers. In this case, you may have drilled down from a broad causal relationship (reading more books helps students perform better on a standardized test) to a purely correlative one (reading fiction is related to higher test scores than reading nonfiction). When doing data analysis for decision support, be careful not to overlook major trends when examining microrelationships. Be certain to look at data holistically on a regular basis.
Go to Step 4 > > >
< < < Return to Intro
Read other articles from the May Issue
Send a letter to the Editor in response to this article.
|
