|  |  |  | Probe |  | Gene |  |
---|
Models | Endpoint | Model Number | Mean (Variance) | Total | Overlapped | Rate(%) | Endpoint Overlap | Total | Overlapped | Rate(%) | Endpoint Overlap |
---|
Normal | D | 28 | 97.04 (91310.26) | 1747 | 345 | 19.75 | 785 | 1350 | 402 | 29.78 | 619 |
 | E | 22 | 143.5 (95148.74) | 1760 | 644 | 36.59 |  | 1309 | 589 | 45.00 |  |
Swap | D | 20 | 54.10 (3964.62) | 609 | 207 | 33.99 | 317 | 465 | 174 | 37.42 | 252 |
 | E | 20 | 106.05 (32074.79) | 1047 | 443 | 42.31 |  | 793 | 416 | 52.46 |  |
- This table shows prime analyses at levels of probes and genes for normal models and swap models. Total refers to the total number of irredundant probes and genes, with means and variances of models for each endpoint be listed in Mean (Variance). Rate refers to the proportion of overlaps in unique sets of probes and genes. The numbers of overlaps between unique sets of probes and genes of both endpoints are calculated, which are displaced in Endpoint Overlap. Note that two probes for NIEHS_BR_E_5 were removed as they do not appear in the Affymetrix U133A platform. We compared lists of probes for each group of data to get overlaps at probe level. If two probes overlapped at probe level, they must also overlap at gene level; if two probes are not same but they share the same gene, they can also overlapped at gene level. This table suggested that the overlap rates of endpoint E are always greater than those of endpoint D. For normal models, the number of features for both endpoints have no significant difference (F = 0.9507, T-test = 0.5748) and there is either no significant difference for that of swap models (F = 0.1236, T-test = 1.2238, degree of freedom = 23.6263). Furthermore, overlap rates at gene level are greater than those at probe level, suggested that the number of non-identical probes which share the same genes is large.