Another and perhaps more natural way to see the effect of LD is to plot the two disease parameters against each other. We refer to this as a model space plot, since each point corresponds to a particular disease model and all possible models can be represented in this way (up to the value of µ). Figure 4 shows such a plot with curves for each of the eight scenarios shown in Figures 2 and 3. The subspace of multiplicative models is shown by the horizontal line, and the null model is at the origin. The curves trace out the theoretical disease model at the marker SNP, with lower LD corresponding to points closer to the origin along these curves. We can now clearly see how LD acts to make the observed model more multiplicative—notice that the curves “bend” toward the horizontal line.
