Fig. 4From: Practical impacts of genomic data “cleaning” on biological discovery using surrogate variable analysisGlobal view of the impact of differing models in SVA. PCA performed on the original normalized data (“No SVA”), and on data “cleaned” with SVA using the 3 different models described in the text (a, b and c). Values of individual samples in the first 4 PCs are shown for each analysis, along with a smoothing spline to these PC’s across age (red). SVA Model A: linear spline with a knot at birth [2° of freedom], SVA Model B: 2nd degree basis spline with a knot at birth [3° of freedom], SVA Model C: 2nd degree basis spline with knots at birth, 1, 10, 20 and 50 years [8° of freedom]. Each model also incorporated an offset at birthBack to article page