The four levels of measurement are nominal, ordinal, interval, and ratio. Nominal measures classify elements into categories of a variable that are exhaustive and mutually exclusive, and represent characteristics or attributes of a variable. They are the lowest level of measurements. There is no rank-order relationship among categories, and the investigator cannot measure the distance between categories. When numbers are assigned to the categories, they mainly serve as codes for the purpose of data collection and analysis, but have no numerical meanings. Ordinal measures refer to those indicators or variables whose attributes may be logically rank-ordered along some progression. They may also be used to allow respondents to rank certain items in terms of their preference. Ordinal

Question

measurement is more advanced than nominal measurement. In addition to the rank-order function, ordinal measurement contains all the characteristics of a nominal measure, including classification, exhaustiveness, and exclusiveness. Interval measures refer to those variables whose attributes are not only rank-ordered but are separated by equal distances. Interval measures are more advanced than either nominal or ordinal measures and contain all the properties of the others. Interval measures do not have a true zero, and zero is merely arbitrary. Because of this arbitrary-zero condition, there are very few interval measures in health services research. Very often, interval measures are mixed together with ratio measures, called interval-ratio measures, and the arbitrary-zero requirement is ignored. Ratio measures are similar to interval measures, except that ratio measures are based on a nonarbitrary or true zero point. The true-zero property of ratio measures makes it possible to divide and multiply numbers meaningfully and thereby form ratios. Like interval measures, ratio measures are a higher level of measurement than nominal and ordinal measures, and contain all their properties.

What will be an ideal response?

Answer 1

Systematic errors refer to certain biases consistently affecting the measurement process. They may be caused by inaccurate operationalization of the concept of interest or because of the research experience itself. Important dimensions or categories of the dimensions of the concept may be missing so that there is only a weak link between the categories used and the concept they try to measure, or too few items may be used to measure a particular concept, making it difficult to assure their accuracy. Observations may be biased when respondents act differently in front of researchers from their usual behavior, or the study may be distorted by low response or high attrition. Systematic errors are serious threats to study validity and reliability.

Nonsystematic errors, or random errors, refer to certain biases affecting the measurement process in haphazard, unpredictable ways. These types of errors occur when characteristics of individuals, respondents, or investigators affect the measurement process. Respondents may have fluctuations in mood, ignore an item or respond spontaneously if they do not know its contents, or may be too tired or distracted to think through the answer. Investigators may use improper interview methods, such as not following instructions, changing the wording or order of questions, asking leading questions, making subjective interpretations of respondents’ answers, or omitting instructions due to fatigue or lack of interest. Both respondents and investigators may be turned off by each other due to their demographics, attitudes, or demeanor, and both are also prone to classification errors. Random errors do not present a great threat to research, but they may affect study reliability.