| Abstract: | The criteria used by program directors in surgery for selection of surgical residents varies widely across programs. While most criteria tends to include one or more interviews, grades, experience, and letters of recommendation, there are very few models to facilitate the selection process, and the existing ones are linear in scope. This paper presents a neural network approach to facilitate the surgical residents selection process using 36 variables. Multi-layer perceptrons employing backpropagation are shown to result in better separation than stepwise discriminant analysis using a paired t-test. |
