Epidemiology, health policy and outcomes
Kaleb Michaud, PhD
University of Nebraska Medical Center
.png "Becki Cleveland, PhD photo")
Becki Cleveland, PhD
University of North Carolina
.png "Jamie Collins, PhD photo")
Jamie Collins, PhD
Brigham and Women's Hospital
Selecting the appropriate statistical analysis is a very common problem for novice and experienced researchers alike. However, knowing which statistical test to use, or whether a statistical test is needed, is critical to producing accurate study results. The correct statistical analysis will depend on a number of factors related to the study design, types of variables and the study question. To determine the correct statistical test, the researcher will need to determine if the data are independent or dependent, and the types of variables for both response and predictors. For continuous variables, it is necessary to determine whether data are normally distributed. For categorical variables, researchers must determine the data are binary, nominal or ordinal. The type of analysis will also depend on the research question. Descriptive analyses--for example, assessing disease prevalence--does not require a hypothesis test. There are also hypotheses of difference and hypotheses of association which will require different statistical analyses. As p-values and hypothesis testing come under scrutiny in the methodological literature, determining whether and when statistical testing is necessary is an important step. This session will help researchers determine the appropriate statistical test based on their study design, types of variables and research question.
