The statistical methods in this program all involve the analysis of relationships between mea surements made on a group of observations. The observations can be anything, patients, blood samples, animals, trees, and so on. For example, the measurements might be the systolic blood pressure of patients, or the milk production of cows under various feeding schemes. We dis tinguish two types of variables, response or outcome variables and explanatory variables. The response or outcome variable is a random variable and the explanatory variables are fixed vari ables that explain or predict the variation in the random variable.
Sometimes the response variable is called a dependent variable and the explantory variables are called independent variables.
Measurement Levels
Nominal or categorical. This is a classification variable, where observations in the same category are identical and different from other categories. Examples are eye colour with categories green, blue, brown. Another example is religion with categories protestant, catholic, Hindu, and other. A binary or dichotomous variable is a special categorical variable that assumes only two values, which are typically coded zero and one. Examples of a binary variable are: dead, alive; male, female; treatment group, placebo group; and so on.
Ordinal. This is a categorical variable, where there is an order or ranking of the categories of the variable. An example is a rating on a diagnostic test, with categories normal, borderline, and abnormal.
Continuous data, where the measurements are made on a continuum. Examples are length, weight, and body mass index.
Count data, where the measurements are counts, here a Poisson regression model could be an appropriate model. Examples are the number of accidents per day, the number of eggs in a nest, and the number of doctor visits. The measurement levels of the response and explanatory variables guide the choice of the ap propriate statistical analysis.