Article ID: | iaor199981 |
Country: | United States |
Volume: | 29 |
Issue: | 6 |
Start Page Number: | 487 |
End Page Number: | 495 |
Publication Date: | Jun 1997 |
Journal: | IIE Transactions |
Authors: | Bailey Charles D., McIntyre Edward V. |
Keywords: | learning curves |
Learning-curve models fitted to initial data are used to predict subsequent performance; however, the model that fits the initial data best may not predict best in future periods – a paradox documented in applications of other prediction models. Little evidence exists about the magnitude of the problem in the domain of learning curves and relearning curves. Using laboratory data, the authors examine the predictive ability of alternative models, examine the strength of the relation between goodness-of-fit and predictive ability, and test whether this relation is the same for both learning curves and relearning curves. Although the correlations between measures of goodness-of-fit and predictive ability are not high, one curve (a log–log–linear model recently introduced to the literature) tended to dominate the rankings on the basis of predictive ability for both learning curves and relearning curves. This curve also tended to provide the best fit in the estimation period as a relearning curve, and the second-best fit as a learning curve.