Table 3.
Parameters and estimated values of the final linear model of the data-subset specific relation of LINE-1 and LUMA assay-based measurements of global DNA methylation. The full model was given as MethylationLINE1 = (Intersection + θ Int, Subset1.. Subset3) + (Slope × θ Slope, Subset1.. Subset3) × MethylationLUMA + ε, where Intersection and Slope are structural parameters of the linear model denoted during the fitting as θ 1 and θ 1, respectively, θ Int,Subset1.. Subset3 and θ Slope, Subset1.. Subset3 are data subset specific modulators of the structural parameter values, denoted during the fitting as θ 3..12, and η 1 and ε accounts for the additive error in the fit of the percent methylation data acquired by means of two different assays. The final model was the result of the model building favoring the best but sparsest model based on goodness-of-fit statistics
| Parameter | Value (and % SEE) | 95 % bootstrap CI |
|---|---|---|
| Y-intersection (% methylation) = θ 1 | 0 (fixed) | – |
| θ 3 = θ Int,Subset1 | 0 (fixed) | – |
| θ 4 = θ Int,Subset2 | 9.03 (9.5) | (7.3, 19.9) |
| θ 5 = θ Int,Subset3 | 71.1 (8.4) | (58.7, 87) |
| Slope = θ 2 | 0.38 (1.1) | (0.32, 0.4) |
| θ 6 = θ Slope,Subset1 | θ 6 = θ7 = 2.68 (0.7) | (2.2, 2.8) |
| θ 7 = θ Slope,Subset2 | SlopeSubset1,2 = 2.68 · 0.38 = 0.9 | |
| θ 8 = θ Slope,Subset3 | 0.88 (27.7) | (0.2, 1.3) |
| SlopeSubset3 = 0.88 · 0.38 = 0.29 |
SEE standard error of parameter estimate, fixed the parameter was not estimated but set at the shown value, CI 95 % confidence interval of parameter estimate, obtained from 1000 model runs of the final model with bootstrap resampled data