There are two recent articles that should be of interest to anyone who teaches statistics. The first is Tom Siegfried’s article in the March 27th (Vol 177 page 26) issue of Science News, “Odds are, It’s Wrong: Science fails to face the shortcomings of statistics.” At the time of the writing of this entry, this article can be retrieved from:
Of a similar vein is an article by Frank Schwartz called “Detecting and Correcting the Lies that Data Tell” in May’s edition of Perspectives in Psychology. For those of you with APS membership or who, through your university, have access to the journal, the article can be retrieved by going to http://pps.sagepub.com/content/5/3/233
It seems that if there is going to be a blog on the teaching of statistics, it seems appropriate to begin by discussing why some of the most commonly used statistics are even useful to the sciences. It seems even more appropriate to discuss the writings of individuals who feel such statistics are detrimental to the sciences.
Please permit me to summarize the main points I see in Siegfried and Schwartz’s article.
(1) Statistics is not fully understood and thus improperly implemented. This is resulting in a negative impact in the sciences and in other disciplines like business and education who rely upon statistics to answer questions.
(2) People throughout the sciences lack even the most basic understanding of psychometrics, that is the measuring of humans. They do not “get” measurement error, and that even the best uses of statistics can’t undo the impact of a lousy measure.
(3) Schwartz, in particular, appropriately encourages the teaching of confidence intervals and effect size statistics. (Topics that will most assuredly arise in later blogs).
It is true, I agree with each of the previous points. However, this is where my thinking diverges from Siegfried’s and Schwartz’s. As it seems Siegfried and Schwartz’s method of dealing with this very real situation is to ditch statistical hypothesis testing, like the t and the F test. People are misusing them, do not understand how they work, and are implementing them at times when other statistics would provide more useful information.
I say, statistical hypothesis testing still has a place. However, in order for the tools of statistics to be properly implemented (no matter if we are discussing CI, t’s, F’s, etc.), students must be properly taught. Coupled in this is not just the teaching of statitics but also of research methodology and psychometrics.
Over the next few weeks, the Sages and I will be discussing how to optimally teach statistical hypothesis testing so students will understand its uses, limitations, and when to make use of other statistics. We will address the issue of measurement error and psychometrics in future months.
For now, we look forward to hearing from you on your thoughts on Siegfried and Schwartz’s position on statistical hypothesis testing.