For many years, I have been a faculty member in the College of Nursing at Michigan State University, where I teach, among other courses, research design and statistics in the Master’s programs to prepare nurse practitioners and nurse anesthetists. I also teach a research methods course in our PhD program preparing future nurse researchers and faculty. Last fall, we started a Doctor of Nursing Practice program, which, while more clinically oriented than the PhD program, will also produce many graduates who will end up teaching in nursing programs.
Given my experience, I am well aware that many students dread quantitative research in general and statistics in particular. At the same time, it is generally accepted in the nursing profession that nursing practice should increasingly be based on empirical evidence. Yet for evidence-based-practice (EBP) to become well established, clinicians, and their teachers, must be able to read the relevant research literature. Much of this literature consists of quantitative research reports from studies that employ complicated research designs and ever more sophisticated statistical analysis techniques. In short, while EBP has become the accepted norm for all clinicians, the gap between researchers and clinicians is becoming wider and wider.
It is my conviction that the future clinician must know more than just a few descriptive statistics and analysis of variance or linear regression. However, even more important than the grasp of additional statistical models encountered in the research literature is an understanding of the underlying logic of these models. First and foremost, the user of statistical information must understand the logic of statistical inference, since a solid grasp of the principles of inference can go a long way towards providing a base from which to expand into newer statistical models, which will undoubtedly be introduced in the clinical literature over the next 20-30 years.
Unfortunately, many undergraduate introductory courses do not emphasize understanding, but often rely on the memorization of formulas, which are forgotten as soon as the course is over. Thus, my colleague Dr. Dontje and I have tried a different approach: we start “from scratch,” not assuming anything about the statistical knowledge of our students. In our new book, Statistics for Advanced Practice Nurses and Health Professionals, we introduce the core ideas of statistical inference through a discussion of the t-test and then rapidly advance towards more sophisticated statistical models. One of our major concerns is to show the similarities among many of the statistical models. For instance, many students do not know that the t-test, analysis of variance, and linear regression rest on the same assumptions and mathematical underpinnings. Understanding such cross-references not only deepens our understanding of statistical models, but actually makes it easier to comprehend them, since they are subspecies of larger models. Similar cross-illuminations occur when discussing odds-ratios in the context of the logistic regression model and relative-risk ratios in the contexts of hazard regression models.
In short, our goal is to provide readers of the clinical research literature with a good basic understanding of commonly encountered statistical models, so that the information is interpreted correctly. There is no easy way to acquire this level of understanding, and we don’t pretend otherwise. It requires one to dig a little bit deeper, but the rewards are incomparably greater: we envision a clinician, who reads a clinical trial report or a report on a cohort study and is able to judge the adequacy of the research design and analysis decisions; at the same time, he or she feels confident in correctly interpreting tables with numerical and statistical information.