Table 3 Logistic regression models for two groups. Logistic regression fits contrasting normal colon with cancer samples for tag ATTTGAGAAG from Table 2. The first model makes no allowance for overdispersion, and the latter two introduce it in different ways. The introduction of overdispersion is important as it dramatically affects the results, but the choice of overdispersion method is less crucial.
From: Overdispersed logistic regression for SAGE: Modelling multiple groups and covariates
Model 1: | No overdispersion | V(Y i ) = n i p i (1 - p i ) | Â | |
|---|---|---|---|---|
Coefficients | Estimate | (s.e) | z-value | p-value |
β 0 | -4.660 | 0.033 | -140.68 | < 2e-16 |
β 1 | -0.888 | 0.043 | -20.41 | < 2e-16 |
Model 2: | Quasilikelihood | V(Y
i
) = n
i
p
i
(1 - p
i
) |
| |
Coefficients | Estimate | (s.e) | t-value | p-value |
β 0 | -4.660 | 0.454 | -10.261 | 5e - 05 |
β 1 | -0.888 | 0.595 | -1.489 | 0.187 |
Model 3: | Hierarchical | V(Y i ) = n i p i (1 - p i ) [1 + (n i - 1)φ] |
| |
Coefficients | Estimate | (s.e) | t-value | p-value |
β 0 | -4.656 | 0.428 | -10.874 | 3.6e - 05 |
β 1 | -0.850 | 0.570 | -1.492 | 0.186 |
= 187.6
= 3.4e - 03