Abstract:
|
Many preclinical investigators lack formal training in experimental design, and have only minimal training in basic analysis methods. Textbooks, teaching, and statisticians themselves have created a monster where over-emphasis on analysis encourages investigators to put too much emphasis on null hypothesis significance testing NHST and none on fundamental statistical principles of experimental design, sampling, and randomization. As a result, overall research quality is poor. Studies tend to be small, biased, and noisy, with a high risk of false positives, and serious errors in statistical methodology. Statistical educators should shift emphasis away from NHST and instead prioritize instruction on the fundamentals of good statistical practice. Furthermore, applied biostatisticians should be alert to the necessity of enquiring into the provenance of the data handed over to them for analysis, especially with respect to design, data sampling, and collection.
|