Nominal scales are designed to classify cases into mutually exclusive and exhaustive categories. Diagnosis is assessed with a nominal scale: we simply tally the number of individuals according to the presence of specific and distinct disease states. Few mathematical operations -- other than an analysis of proportions -- can be performed using nominal data: "mean" disease, for example, is not a meaningful calculation. Ordinal data are derived from scales designed to assess variations in "more" or "less" of an attribute, and the distance between items on the scale are non--equivalent. The most recent college sports poll is a representative ordinal rating; one would be pressed to argue that the 10^th rated team is ten times worse than the 1^st place team. Data gathered through use of nominal and ordinal scales are appropriately analyzed with non-parametric statistics, i.e. those that do not depend on population estimates such as the mean or standard deviation.

In interval measurement the distance between measured units is meaningful and so basic addition and subtraction are permissible. But the scale is centered on a relative standard and statements about the fractional change between scores are strained. For example, the Fahrenheit scale is centered on the temperature at which water freezes; and although the distance between units on the Fahrenheit scale reflects an equivalent change in mercury levels, 100^o feels much more than twice as hot as 50^o. Ratio scales, by rule, have the properties of interval scales and a zero point. Physical variables such as weight and length as well as many other variables that can be counted (e.g. dollars earned, clients seen, prescriptions filled) are measured with ratio scales.

Question 17.2

A visual analogue scale is a

	Nominal scale
	Ordinal scale
	Interval scale
	Ratio scale
	Good question

Page 28 of 47
Previous Page

Next Section