The work on my master’s thesis involved looking at offender risk scores over time to see if changes in risk score predict changes in offender risk. The previous research had suggested that predictive validity improves from the first to the second assessment. It had been proposed that the reason that the reason that the second assessment was better at predicting risk was because offenders had changed between assessments and the second assessment was a better predictor because it had captured the change in risk level over time. In the work on my thesis, I demonstrated that the increase in predictive validity only occurred between the 1st and 2nd assessments. Predictive validity did not improve between the 2nd and 3rd, or the 3rd and 4th assessments. Why?
In trying to find the answer to that question, I traveled down a rabbit hole on a ten year odyssey. I ended up spending a million dollars studying the nature of changes in offender risk over time. I suppose that I should have published my work as I went along, but I was not satisfied that I understood what was happening. By the time I had figured things out to my satisfaction, I was running out of funds. I decided to go back to work and finish this later.
Here is the thesis. As a side note, it won an award for being the best written master’s these at St. Cloud State University in 2007. Spoiler alert: I eventually figured out why I found what I did.
Dynamic Changes in the Level of Service Inventory-Revised and the Effects on Prediction Accuracy