(a) Rank of the relative importance for each parameter (-axis) across all three analyses (-axis). Warmer colors indicate parameters that have a greater influence on statistical estimates. (b) Statistical distributions from the three analyses, shown as kernel density plots, separated by choice of gene normalization method (the most impactful parameter as shown in panel a). (c) Density plots of the statistical estimates for all 746,496 pipelines shown along the first two principal components, derived from the 746,496 (pipeline) x 3 (statistical estimates) matrix, representing how different the statistical estimates from each of the three analyses are relative to other pipelines. Left panel: pipelines are colored based on choice of gene normalization method, where each color represents 1/3 of the pipelines. Here, the pipelines in which no normalization was applied (purple) are distinguished from those in which some form of normalization was applied (blue and brown). Right panel: pipelines are colored based on whether gene normalization was performed within (True, red) or across (False, purple) structural classes (i.e. cortex, subcortex/brainstem, cerebellum; see Materials and methods: Gene expression pipelines for more information).