We recommend the following if LD between markers is a concern in the data. Pick a small integer k > 0, corresponding to the number of adjacent markers one uses for adjustment (k = 1 will often suffice). In the data matrix M we will “predict” each column by running a multivariate regression on the k previous columns. We then will analyze the residuals. Concretely: we first form M, as in Equation 2. For each column j
