Data fabrication can occur at various stages of toxicological research. This includes initial data collection, where researchers might invent results, or during data analysis, where they might manipulate datasets to fit desired outcomes. It can also happen through the selective reporting of results, often termed "publication bias," where only favorable results are reported, ignoring unfavorable or null outcomes.
