Figure 17
Five ordinary STIs for the (a) real and (b) simulated mac1 datasets. The line of equivalence (black) has slope 1 and intercept 0 which corresponds to the case of Cy5 = Cy3. (c,d) Scatter plots of residuals for the real mac1 dataset versus predictor variable. On this figure (c,d) and other figures 'residuals' mean 'log2(Cy5/Cy3)'. The residuals are depicted in (c) and this is the same as (a) except that the linear trend has been subtracted resulting in a slope A = 0. The ordinary STIs assume residual homoscedasticity. (d) STIs shown corrected with the S-plus scatter plot smoother supsmu to reveal the dependence of residual variance on the value of the predictor variable. The supsmu-based STIs assume residual heteroscedasticity. Pink and cyan points lie in the interval between the upper and lower 95% and 99% STIs, respectively. Red and blue points lie above the upper and lower 99% STI, respectively. Black points lie below the upper and lower 95% STI. In all panels, the 95% (innermost), 99%, 99.8%, 99.98% and 99.998% (outermost) STIs are shown (covered with probability at least 0.9999). The vertical dotted line marks the location of the minima of the empirical hyperbolas. Therefore, red/pink, blue/cyan and black points represent up-regulated, down-regulated and unchanged genes, respectively.