Tareq Kheirbek MD ScM FACS
There are two major types of data: Categorical and Continuous. Your test first depends on what kind of data is in your outcome variable. Categorical data can be either binary (dichotomous, most commonly used), or several categories (nominal or ordinal)
Variable is Categorical: there are separate independent classes of outcomes:
- Two possible outcomes: binary (yes or no, alive or dead, operative or nonoperative)
- Three or more possible outcomes but in random order: nominal (discharge destination: home, rehab, transfer, death)
- Three or more possible outcomes but in a ranked order: ordinal (easy, moderate, severe ARDS)
In categorical data, especially binary, it is important to determine if the categories are independent (unpaired) or dependent (paired). Most of your data will be independent. Example of dependent categories is measurement of BP in individuals before and after an intervention, since you will be comparing the BPs in each individual so they readings before and after are paired.
Variable is continuous: there are indefinite options of outcomes that are on a continuous scale (age, los, temperature, etc). While some variables may be documented in an ordinal fashion (integers) if conceptually the outcome would accept fractions of the integers as values then it should be counted as continuous.
Question for discussion: How to define ISS?
For continuous exposure data, we next determine whether the data is normally distributed or skewed (right or left). This will help us determine whether to use a parametric or nonparametric statistical test.
