Figure 1

Comparison of PIR and SAM regression methods. Panels A & B plot the penalized isotonic regression (PIR) test statistic (M _i , x-axis) against the penalized linear regression (SAM) test statistic (d _i , y-axis) for deltamethrin and permethrin, respectively. All 31,042 probe sets present on the Affymetrix Rat 230 2.0 GeneChip^® are shown. Data points in green have an empirical p-value < 0.01 for both the PIR and SAM methods. Data points in blue have an empirical p-value < 0.01 for the PIR regression only. Data points in orange have an empirical p-value < 0.01 for the SAM regression only. In the deltamethrin and permethrin analyses, 49.5% and 53.7% of all probe sets identified by either the PIR or SAM method had p < 0.01 for both methods. Data points circled in red have a q-value < 0.10 in permutation-based FDR calculations employed in the SAM algorithm. Note that the rank order of statistical significance was similar between the two methods in that probe sets commonly identified using the PIR or SAM method tend to appear in the upper-right and lower left hand corners of the scatterplots (green points).