Statistics is No “Bloodless” Algebra

Abstract

Because biological processes are full of variations, statistics can not give you certainties, but only chances. What kind of chances? Basically, the chances that prior hypotheses are true or untrue. The human brain excels in making hypotheses. We make hypotheses all the time, but they may be untrue. E.g., when you were a kid, you thought that only girls could become doctors, because your family doctor was a girl. Later on this hypothesis appeared to be untrue. In clinical medicine we currently emphasize that hypotheses may be equally untrue and must be assessed prospectively with hard data. That’s where statistics comes in, and that is where at the same time many a clinician starts to become nervous, loses his/her self-confidence, and is more than willing to leave his/her data to the statistical consultant who subsequently runs the data through a whole series of statistical tests of SAS¹ or SPSS² or comparable statistical computer software to see if there are any significances. The current article was written to emphasize that the above scenario of analyzing clinical trial data is bad practice and frequently kills the data, and that biostatistics can do more for you than provide you with a host of irrelevant p-values.