The number of samples n needed to get an overlap f ≥ fc = 1 − ε, with confidence 1 − δ, for two PGLs of size αNg
| fc = 1 − ε | δ | n (10) | n (11) |
|---|---|---|---|
| 0.02 | 0.5 | 87 | 104 |
| 0.05 | 0.5 | 170 | 218 |
| 0.10 | 0.5 | 290 | 383 |
| 0.20 | 0.5 | 553 | 743 |
| 0.50 | 0.5 | 2,300 | 3,142 |
| 0.02 | 0.1 | 178 | 195 |
| 0.05 | 0.1 | 270 | 319 |
| 0.10 | 0.1 | 412 | 507 |
| 0.20 | 0.1 | 736 | 930 |
| 0.50 | 0.1 | 3,026 | 3,883 |
-
We use α = 0.0046 (corresponding to a PGL of 70 genes) and α = 0.0068 (corresponding to a PGL of 76 genes) for refs. 10 and 11, respectively. For δ = 0.5, fc = f*n , and hence n represents the number of samples needed for an average overlap of 1 − ε. The effective number of genes used here (after preprocessing) was Ng = 15,125 for ref. 10 and Ng = 11,130 for ref. 11.
