Table 1.
Fitted and tabulated classification rates (percentages) reflecting only reading error.
| True stage u: | Estimated percentage in each read stage, given each true stage* |
|||||||||||||||
|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|
| 0 | 1 | 2 | 3 | 4 | ||||||||||||
| Read stage v: | 0 | 1 | 0 | 1 | 2 | 1 | 2 | 3 | 1 | 2 | 3 | 4 | 3 | 4 | ||
| Specimen effects† | β | |||||||||||||||
| No | −1 | 93 | 7 | 15 | 79 | 6 | 20 | 77 | 2 | 6 | 40 | 52 | 1 | 16 | 84 | |
| 0 | 83 | 17 | 5 | 78 | 16 | 8 | 85 | 7 | 3 | 20 | 72 | 5 | 6 | 94 | ||
| +1 | 65 | 35 | 2 | 63 | 36 | 3 | 79 | 18 | 1 | 8 | 76 | 15 | 2 | 98 | ||
| m‡ | 81 | 19 | 7 | 73 | 19 | 10 | 80 | 9 | 3 | 23 | 67 | 7 | 8 | 92 | ||
| Yes | −1 | 86 | 14 | 24 | 63 | 13 | 29 | 64 | 7 | 6 | 41 | 48 | 5 | 25 | 75 | |
| 0 | 74 | 26 | 12 | 62 | 26 | 16 | 69 | 15 | 4 | 28 | 56 | 12 | 13 | 87 | ||
| +1 | 60 | 40 | 5 | 54 | 41 | 7 | 65 | 27 | 2 | 16 | 58 | 24 | 6 | 94 | ||
| m‡ | 74 | 26 | 14 | 60 | 27 | 17 | 66 | 16 | 4 | 28 | 54 | 14 | 15 | 85 | ||
| Raw tabulations across all readers and specimens: | ||||||||||||||||
| Study R1 | 82 | 18 | 6 | 75 | 18 | 9 | 82 | 8 | 3 | 22 | 69 | 6 | 7 | 93 | ||
| Study R2** | 47 | 50 | 3 | 12 | 74 | 9 | 6 | 15 | 59 | 21 | ||||||
*
Values shown are 100× r̂j·uv as defined at the beginning of Section 3.1, where j corresponds to a reader with the β shown for that row, and the · subscript indicates averaging over specimen effects if present. Bold entries are the percentage correctly classified for each true stage.
†
If present, specimen effects are assumed to be −1.92, 0, and +1.92 each with probability 1/3.
‡
Marginal rates, averaged over readers with β= −1, β=0, and β= +1.
**
Two observations (6%) with true stage 2 and read stage 0 not shown due to space constraints and no occurrence of this combination in Study R1.