Table 2. Mean absolute errors (MAEs) for models trained and applied on different age ranges.
MAE by Age Bin | ||||
Age Group Tested | ||||
Model Trained On | [1-18) Years | [18-45) Years | [45-65) Years | 65+ Years |
Top-5[1,18) | 1.35 | 7.17 | 5.11 | 4.88 |
Top-5[18,45) | 1.60 | 6.41 | 5.12 | 5.26 |
Top-5[45, 65) | 2.21 | 7.18 | 3.28 | NA |
Top-565+ | 1.94 | 7.02 | 4.96 | 4.63 |
Top-10[1,18) | 1.16 | 6.36 | 4.79 | 4.61 |
Top-10[18,45) | 1.33 | 5.51 | 3.19 | 5.00 |
Top-10[45, 65) | 2.12 | 6.80 | 2.91 | NA |
Top-1065+ | 1.53 | 6.57 | 4.79 | 4.49 |
Models containing only the Top-5 and Top-10 variables were trained in one age group and tested on all age groups. MAEs are from testing on the 20% held-out dataset in each case.