A statistics primer. Statistical tests for discrete data.

When reviewing articles in the literature, it is important to spend some time thinking specifically about the data presented. The data collected in an experiment may be classified into two general types. Continuous data represent data that vary over a continuous range and may be displayed on a curve, frequently a bell-shaped curve. Some examples of continuous data include height, age, and the times recorded in a mile race. Discrete data, on the other hand, are data that fall into specific categories. Examples of this type of data include gender, the presence or absence of a specific condition, and the classification of a patient’s results into groups such as poor, fair, good, and excellent. Statistics provides different tests for analyzing these different types of data. In this review some of the commonly used tests for analyzing discrete data will be presented. The frequency with which an observation occurs may be evaluated with the chi-square test. Suppose you read an article that reviews the frequency of three types of acromion shapes (flat, curved, or hooked) in 33 cadaveric specimens with rotator cuff tears. The null hypothesis would state that the acromion shape has no relation to rotator cuff tears and the expected or theoretical frequency of specimens with rotator cuff tears should be the same with each type of acromion; that is, 11 specimens should be found in each of the 3 groups. The data that were collected, however, show a higher incidence of rotator cuff tears in specimens with curved acromiae, and an even higher incidence in specimens with hooked acromion types, with 3 specimens with flat acromiae, 8 with curved acromiae, and 22 with hooked acromiae.