Table 2. Definitions of the mathematical quantities used.

Item	Definition
y	The vector of pixel values transformed using the Anscombe transform. The value of the ith pixel is given by yi.
X	The design matrix describing the linear model. A row in the design matrix is given as x i; each row gives the explanatory variables for pixel i.
β	The vector of model parameters which are estimated from the quasi-likelihood algorithm. The jth parameter is given by βj.
μ i	The estimated pixel values. These depend on the value of the parameters as μ = Xβ. The pixel value estimate for the ith pixel is given by μi.
ri	The residual for the ith pixel given by ri = (yi − μi)/vi 1/2.
W	The diagonal matrix of weights, where the ith diagonal element is given by w(ri), the weight for the ith residual.
ρ(ri)	The robust function of residuals.