Figure 3

Comparison of spike log-odds obtained for a particular contrast in the linear model fitted to the 8 arrays involved in that contrast. On the left we show the difference between the log-odds obtained after VST and the log-odds obtained after a log₂ transformation. On the right, we show the difference between VST and a linear model incorporating log₂ variances as weights (see [7]). In the top panels, we show six independent contrasts with the closest spike concentrations. The bottom panel shows six independent contrasts chosen to provide a range in anticipated log-ratios (the finer differences being to the right of the panel). In all cases, a positive value indicates greater log-odds obtained (i.e. more evidence for differential expression) after VST.