A brief overview is given of the types of statistical tests that are available to analyze pharmacy research data.Summary
The most important aspect of selecting the correct statistical test is defining the types of variables being analyzed. Variables that are controlled or determined by the researcher are referred to as independent variables. Dependent variables are those that are observed and are out of the researcher's control. There are two types of random error that exist with inferential statistics: rejecting a null hypothesis (H0) when it is true and failing to reject H0 when it is false. There are two primary ways to interpret the significance of results from an inferential statistical test: (1) creation of a confidence interval and determination of whether a value falls within the interval and (2) calculation of a ratio and determination of whether the resultant value exceeds an established critical value. Student's t test is one of the simplest inferential tests and can be used to illustrate both the confidence interval and the ratio approaches to evaluating sample data. The p value indicates the amount of error that can exist if the researcher chooses to reject H0. Parametric tests require two additional assumptions in order to be applied correctly. Some examples of these include the two-sample t test and the paired t test. Nonparametric tests are designed for small sample sizes and are easy to calculate. These tests use the median as the measure of center. Some examples of nonparametric tests include the chi-square test and the Fisher exact test. Other statistical tests that are available to help the pharmacist researcher include equivalency testing, survival statistics, and noninferiority studies.Conclusion
Selection of the proper statistical test depends on the type and number of variables and whether parametric conditions are met.