The function of proteins in complex cellular processes is regulated by their post-translational modifications (PTMs), and a comprehensive understanding of PTMs is crucial for the emerging field of proteogenomics. Affinity enrichment followed by liquid chromatography coupled with mass spectrometry (LC-MS) is currently the most successful workflow to systematically identify and quantify relative changes for PTMs with great throughput and depth. Through LC-MS, multiple modified and unmodified peptides that span a PTM site can be identified and quantified. However, to characterize the relative change of a PTM site between experiment groups, specialized statistical methods are required to deal with sparsity and confounding changes in overall protein abundance and enrichment efficiency. We present the application of linear mixed-effects models for characterization of quantitative changes in PTMs in global proteomics experiments. Using benchmark datasets from controlled mixtures, we demonstrate that the proposed approach improves the estimation and detection of differentially modified PTM sites. Our approach is implemented in MSstatsPTM, an open source R software package.