Skip to main content
. 2015 Apr 17;16(6):950–963. doi: 10.1093/bib/bbv019

Figure 3.

Figure 3.

Variance comparison. (A) Boxplots of the variance distribution. The variance of the log2 counts of all non-spike-in miRNAs was computed across the samples, and visualized using boxplots. The data are grouped according to the normalization method. Values beyond the 1.5 interquartile range (IQR) are not shown. A clear increase in variance is observed in the data normalized by cpm or total count scaling, while a decrease is seen in UQ or TMM normalized data. (B) Mean-variance dependency. The relationship between variance and abundance level is visualized by plotting the ratio of the variance between the normalized and unnormalized data versus the average counts. Data points below the y = 0 line have decreased variance compared with the raw data. The Lowess smoother line reveals the presence of any trends in the data. A colour version of this figure is available at BIB online: http://bib.oxfordjournals.org.