Fig. 3.
Effect of correlation between variables. This figure shows the results of 100 simulations under each of H0 and H1 for binomial data with N = 100,K = 200 and P = 3450 when the variables are intercorrelated according to an AR(1) model with lag 1 correlations 0.5. (a) No correction for correlation. (b) Correction using estimated inverse correlation matrix from a sample of size 200. (c) Correction using a sample of size 500. (d) Correction using a sample of size 1000.