The Development of an Officer Training School Board Score Prediction Method Using a Multi-Board Approach

One phase in the Officer Training School OTS selection process is meeting a selection board. Recruiting Services HQ ATCRS currently uses a regression model based on data from one board to evenly distribute quality among the different panels in the selection boards. HQ ATCRS wanted a method of predicting board scores based on data from multiple boards. This study used the results from eighteen boards and 9215 applicants to develop and validate a multi-board regression model for each of five application categories. Comparisons between the two models showed mixed results. In two categories, non- rated operations and technical, the single-board model explained a higher proportion of the total variance. However, in the other three categories, the multi-board model explained significantly more of the total variance. In all but one category, the single-board model and lower prediction errors. Overall, the multi-board model was able to predict board scores well enough to sort the records so that each panel would get approximately the same quality distribution of records. A discriminant analysis was also performed using the top 33 of the board scores to represent the records that would be selected. The results showed that the model could not successfully identify the records that would be selected.