Comparing the accuracies of entire-group and subgroup models to predict NBME-I scores for medical school applicants
| Autor: | Mohammadreza Hojat, J J Veloski, J B Erdmann |
|---|---|
| Rok vydání: | 1992 |
| Předmět: |
Adult
Male medicine.medical_specialty Group (mathematics) business.industry Medical school General Medicine Entrance exam Education Age and gender Physical therapy medicine Educational Status Humans Female Prediction bias Educational Measurement business Predictive modelling Schools Medical Clinical psychology |
| Zdroj: | Academic medicine : journal of the Association of American Medical Colleges. 67(12) |
| ISSN: | 1040-2446 |
| Popis: | To address the question of whether prediction models for subgroups of medical school applicants lead to more accurate predictions of performance than does one model for an entire group of applicants, the authors used data from two groups of students at Jefferson Medical College: 415 students who entered Jefferson in 1985 and 1986, and 396 who entered in 1987 and 1988. Both groups were divided into two subgroups by gender and two subgroups by age. Data from the first group were used to develop prediction models based on the entire group and on its four subgroups. The predictors were undergraduate grade-point averages and Medical College Admission Test scores; the criterion measures were scores on the National Board of Medical Examiners Part I examinations. The prediction models were then applied to data from the second group and its four subgroups: differences in the validity coefficients (.40 to .56) and residual scores (7.2 to 17.9) were not considered to be of practical importance. Hence, the authors suggest that gender and age do not contribute to a prediction bias and that an entire-group prediction model can be used without serious concern for over-or underestimating the predicted scores. |
| Databáze: | OpenAIRE |
| Externí odkaz: |
|