| Sumario: | The article presents information on coding reliability. This paper describes a procedure that may help increase the reliability of a coding operation. By using the Random-Systematic-Error coefficient, the survey administrator may be able to determine in cases of low reliability, whether the fault lies in the code or in the coders. Coding disagreements may stem from two sources, the code and the coders. If the categories of a code are poorly defined, overlapping, ambiguous, or inexhaustive of all responses, coders may disagree because there is no dearly correct category for some of the responses. On the other hand, because of personality differences or differential training, one coder may assign one subset of responses to one category while a different coder assigns it to another. Disagreements between two coders using the same code on the same set of responses may be analyzed to indicate their source and the corrective measures to be taken. Errors resulting from a defective code seem generally to be scattered about the range of possible disagreements, while errors originating in the coders tend to fall into systematic patterns.
|
|---|