Search
Presentation Details
| A new set of guidelines to reduce statistical errors in scientific publications. |
| Leonardo B. Castilho. Universidade de Bras�lia, Bras�lia, DF, Brazil |
Statistics has been an essential part of the majority of scientific studies for decades. However, statistics is a science on its own right and its use does not come without challenges. One of such challenges is the need to follow every assumption of any statistical test performed. Fail to do so can lead to untrustworthy results, which hinders the progress of science. Here, I re-analyzed data on 33 papers published in the field of animal behavior in the last decade, and searched for assumptions that were not followed by the original authors. I find that the majority (73%) of the papers published had at least one assumption not met, and almost half (42%), had at least one analysis yielding a conclusion qualitatively wrong, when compared to an analysis in which all assumptions were met. In this talk, I will give a new set of guidelines to correct such worrisome scenario, not only in animal behavior studies, but in every scientific field that relies heavily on statistics. |