Here I refer EWAS to epigenome-wide association study which investigates the association between DNA methylation level (5mC methylation measured by Illumina EPIC array, for example) and some phenotype(s). In the literature, methylation level could be treated as independent variable (x) or dependent variable (y). Here is an article where methylation is treated as the dependent variable in linear regression and the independent variable in mixed linear model. The implication of treating methylation level as the dependent variable or independent variable is usually not discussed in details.
Certainly the choice depends on the research question to be answered, as unlike germline genotypes, methylation levels are dynamic and subjected to influences from various sources. However, for a beginner, a guideline of EWAS would be appreciated.
Therefore my question is: when should methylation level be treated as dependent variable or independent variable? Or put it in other words, what sort of research question could be answered when we treat methylation level as dependent (or independent) variable?