Reproducibility of scientific research has become a major concern in recent years. Inappropriate application of statistical analysis methods by researchers is considered one of the causes of the reproducibility crisis. Drawing on the concept of "statistical myths" – persistent, recurring distortions of statistical concepts – it is argued that these myths are systematic rather than accidental, serve specific functions, and reflect a deep methodological gap between the research domain and the practice of statistical analysis. The relationship between scientific and statistical inference is shown in the form of a biomedical research framework consisting of seven stages: conceptualization and formulation of the research question; operationalization; selection of the estimand; statistical model building and application; interpretation of the estimate accounting for uncertainty; clinical interpretation; and answering the research question. The framework is illustrated using examples from three real-world studies across various medical fields. Based on their functions, the authors classify statistical myths into four categories: epistemological, fundamental, instrumental, and terminological. The defining characteristics and examples of each type are provided, alongside an analysis of several widespread myths. The proposed approach allows statistical myths to be viewed not merely as a pedagogical issue, but as indicators of deeper disruptions in research logic. From a methodological perspective, rather than memorizing yet another list of "correct" and "incorrect" statistical practices, researchers are encouraged to proceed systematically from the research question at hand to its statistical formalization.