Table 4 Expanding contrasts from two to three groups. Logistic regression models testing the significance of a difference between normal colon and primary tumor (β₁) for tag TGCTGCCTGT from Table 2. In the first model, only data from the four libraries directly involved are used. In the second model, data from the four cell line libraries are also included, providing a more stable estimate of the overdispersion parameter φ.

Coefficients

β ₀

β ₁

Model 2:

Coefficients

β ₀

β ₁

β ₂