In math most people publish their papers on arxiv regardless of whether they get published in a journal. Arxiv has its own moderation structure, but theoretically published peer reviewed articles should be more trustworthy because they’re reviewed by your peers. In reality reviewers don’t have time to read papers super closely so some shoddy research gets through even when no blatant corruption is involved. In pure math this isn’t a huge issue because the work usually speaks for itself, but for some areas, especially applied statistics, it’s not so obvious whether an argument is actually well supported or statistically cherrypicked.