July 10th, 2009
So, you’re a big school district or university and you want to put some initiatives in place to help increase the number of students interested in STEM careers. But what changes will have the biggest effect? Will some changes have negative effects you haven’t thought about yet? If you change multiple things, what would be the likely result?
Enter the STEM Research and Modeling Network unveiled today. (Thanks to Curriculum Matters for pointing me to it.) Here’s the nutshell: K-12 districts and higher ed institutions can use this simulation model to enter in possible variables (e.g., quality of math teachers) they would like to change and the model will predict how those initiatives are likely to impact the numbers of STEM-interested students in the pipeline. The example they are leading with is showing how if they enter California’s class size reduction plan into the model, it will predict a shortage of STEM teachers and therefore the lack of impact on students in the STEM pipeline. The authors have over 200 variables in the system. AND they are making it open source so other researchers can contribute.
So, I was of course curious about what is “under the hood.” How is this all working? This document gives the proceedings of a meeting earlier this year where the initial model was discussed. They use what (in system dynamics modeling) is called a stock and flow model. System dynamics is a way of looking at the interrelationships between components of a system that define how the system as a whole behaves. (Oh! They are all connected! Where have I heard that before?) In other words, how do all the various factors interact to affect the flow of STEM-interested and STEM-uninterested students? They first took a lot of existing research studies (listed on their site) that estimate the effects of different variables on STEM interest and then created a statistical model combining all of those studies. As the paper states:
“Raytheon’s model includes 14 validated dynamic hypotheses and 227 independent variables for influencing the number of students who are capable and interested in pursuing careers in STEM disciplines. Model developers used published research on teacher quality and on STEM persistence in college to develop and test these dynamic hypotheses, identify the variables that influence STEM persistence, and determine the relationships between those variables.”
There is a graphical representation of the model in the document.
Here are some things that were raised in the proceedings, and it’s not clear whether have been addressed.
So, there is certainly room for improvement, but I think this is a fantastic start. This is not a methodology that has been used widely in education, so it is interesting in that sense. Given the number of partners involved in creating it, I’m not sure how replicable it is. I’m also not clear yet exactly what statistical methods they are using to model all these relationships. But this ability for stakeholders to find out what happens when they “pull the lever” of one variable or another is very powerful. Of course, it’s only as good as the model, which is only as good as the data that feeds it…
Tags: statistical modeling, STEM interest
Posted in Research Methods, Students | Comments (0)