Distributions of data obtained from the StanfordMicroarray Database. . These probability plots  show wide ranges of linearity, proving the appropriateness of the three-parameter lognormal model (red) for normalizing microarray data. The plots for the original intensity data are also shown (blue). Local background-subtracted  data (green) of ID 22592 could not be normalized since a large proportion of the data (over 25%) was negative; the parameter σ, which is found from interquartile range, for those grids could therefore not be calculated.