Fig. 3
From: Detection of suspicious interactions of spiking covariates in methylation data
Top 6 of the reverse positive or negation CpG sites of the data set by Richter et al. (2019) [17]. The order was determined by the effect deviation between the coefficients from two linear regressions, one associated with all data included and one with exclusion of the spike associated data. In contrast to E-GEOD-54643 and E-GEOD-55454, the spike effect is very marginal. Both regression lines have nearly the same slope with the same negative effect. It is clear to see that an outlier is causing the suspicious interactions