Statistics in Nursing Research

Hayat, Matthew J. PhD

Nursing Research:
doi: 10.1097/NNR.0b013e318257f5dc
Author Information

Assistant Professor and Biostatistician

College of Nursing, Rutgers University

Newark, New Jersey

The author has no funding or conflicts of interest to disclose.

Article Outline

Welcome to this focused issue on statistics in nursing research! Nursing Research put out a special call for papers on this topic in 2011. The response was significant and included 22 submitted manuscripts. Each submission went through a rigorous peer review process that included multiple statistical reviewers. This special issue presents 10 (of 11) accepted manuscripts. The articles cover a broad spectrum of statistical topics and represent recent advances in several important areas of applied statistics, including measurement, multilevel modeling, statistical inference, and causal inference.

Four articles focus on statistical considerations in instrument development and measurement. Stump et al. and Vessey et al. outline advantages to item response theory otherwise missed in classic testing theory. Sousa et al. and Pawlowitz et al. address the growing need in nursing science to compare psychometric instruments. The use of Bayesian statistical methods continues to grow by leaps and bounds in the statistics literature; it is fantastic to see Pawlowitz et al. apply this inference paradigm to a nursing science application. Analysis of correlated data is a hot topic in the nursing sciences, including topics such as the individual health trajectory, multilevel and hierarchical modeling, longitudinal data analysis, and latent variable modeling. Four articles touch on various aspects of these advanced statistical techniques. Hayat and Hedlin provide a nontechnical overview of this topic. Son et al. present an exciting approach for handling missing data using pattern mixture models in SPSS. Schmiege et al. give examples of research questions and corresponding types of latent variable and mixture models that can be used to address them. Eckardt as well as Coffman and Kugler each present methods for making causal inferences using propensity scoring. This is a promising area in the health sciences, as it enables a researcher to make causal statements in situations wherein it was not previously possible. Saylor et al. offer helpful insights into using SPSS to analyze large national databases with sample data obtained with complex sampling techniques.

Let’s consider these topics more generally in the context of science. The lifeline of a scientific investigation begins with a research question. A study is then designed to provide an answer. Rigorous science necessitates careful consideration of statistical considerations in the design of a research study to ensure an adequate answer. The natural next step is to conceive of measurable outcomes and accompanying instrumentation. An abundance of statistical considerations arise in the assessment of instrumentation or measurement. Finally, once data collection is complete, statistical analysis can be performed. The lifeline of scientific investigation is sequential and cumulative. For example, if the research question is poorly formulated, the steps to follow cannot compensate. Similarly, if the instrumentation is poor and does not adequately measure the outcome of interest, a sophisticated data analysis will be of little value.

As the nurse scientist progresses through this cumulative process, it is essential to develop and cultivate a statistical mind-set in the planning stages of research. This needed statistical mind-set goes well beyond considerations of data analysis or thinking of statistics as a tool. It is not enough to know which statistical test to use. Statistical thinking and reasoning is needed. As a statistician in a nursing college and collaborating with nurse scientists, I tell my colleagues to “come early, come often.” Always weary of the garbage in, garbage out principle, my contribution is most significant in the planning stages of research. There is little I am able to do if a research design is poor or instrumentation is inadequate.

Statistical training to develop a statistical mind-set is essential for the nurse scientist. Much can be learned about statistical training from a relatively new subfield of statistics referred to as statistics education research. The need for change in the way statistics is taught and thought about has been long overdue. For example, a study by Hogg (1991) found “Students often consider statistics as the ‘worst’ course they take while in college.” Statistical training for the nurse scientist can and should be fun and interesting. Recent research suggests that before delving into educating students about statistical methods, statistical literacy, thinking, and reasoning training is needed. In fact, studies have shown that a knowledge base in statistical literacy, reasoning, and thinking is needed for understanding published research (delMas, Garfield, Ooms, & Chance, 2007).

I am happy to share news of recent efforts by statisticians to share nursing science with members of the statistics community. “Statistics in Nursing Research” was the theme of a contributed topic session and roundtable discussion at the 2011 Joint Statistical Meetings (JSM, 2011; The Joint Statistical Meetings is the largest gathering of statisticians held in North America. It is held jointly with the American Statistical Association and several international organizations. The meeting was attended by more than 5,000 statisticians. A group of statisticians from nursing schools participated in the conference and contributed to these sessions. In our focused sessions on “Statistics in Nursing Research,” speakers gave talks on statistical applications of multilevel modeling in the nursing sciences, analyzing longitudinal data in nursing research, and experiences and lessons learned in collaborating with nurse scientists. Roundtable discussion included topics related to (a) a faculty appointment as a statistician in a nursing school, (b) promotion, (c) statistics education for nursing students, (d) use of statistical software, (e) statistical consulting, and (f) the interesting multitude of roles for a statistician in an interdisciplinary setting, such as consulting, teaching, research, and service. It was a fun and fascinating experience to connect with other statisticians also collaborating with nurse scientists. There was much interest in continued discussion. We have a contributed topic session, panel discussion, and roundtable discussion on the topic of the use of statistics in the nursing sciences scheduled for the 2012 JSM (

Matthew J. Hayat, PhD

Assistant Professor and Biostatistician

College of Nursing, Rutgers University

Newark, New Jersey

Back to Top | Article Outline


delMas R., Garfield J., Ooms A., Chance B. (2007). Assessing students’ conceptual understanding after a first course in statistics. Statistics Education Research Journal, 6 (2), 28–58.
Hogg R. V. (1991). Statistical education: Improvements are badly needed. The American Statistician, 45, 342–343.
© 2012 Lippincott Williams & Wilkins, Inc.