Cohen's Coefficient Research Articles

Poor agreement between reported and recorded nocturnal cough in asthma.

Reported presence or absence of night cough was compared with tape recorded cough in 15 children with perennial asthma (median age, 9 years; range, 7-14) who reported troublesome nocturnal symptoms. Measurements were made and diaries kept for 7 consecutive nights. Cough was reported on 66 of 105 (66%) and recorded on 93 (90%) available nights with poor overall agreement (Cohen's coefficient of assessment, kappa +0.30, range -0.17 to +1). The poor agreement between subjective and objective assessment of an important symptom of nocturnal asthma raises questions on the validity of symptom reporting and may in part explain the not infrequent disagreement between medical and patient assessment of disease severity.

The use of the coefficient of agreement and likelihood ratio test to examine the development of the tympanic plate using a known-age sample of fetal and infant skeletons.

Previous attempts to use the development of the tympanic plate as an aging standard for fetal and infant cranial remains have either produced only a developmental sequence (Anderson, 1960), or a sequence for which age has been estimated using another independent aging standard (Weaver, 1979). Following an evaluation of the first three stages of development of the tympanic plate using a known-age-sample of fetal and infant skeletal remains, a simple test of interobserver agreement, Cohen's coefficient of agreement (Cohen, 1960), is demonstrated using the developmental stages.

AN INDEX OF DEPENDABILITY FOR MASTERY TESTS

Glaser and Nitko (1971) define a as that is deliberately constructed to yield measurements that are directly interpretable in terms of performance standards (p. 653). This is probably the best-known definition of a but others have been proposed (e.g., Harris and Stewart, 1971; Ivens, 1970; Kriewall, 1969; and Livingston, 1972). Nothing in the Glaser and Nitko definition, or in most other definitions of criterion-referenced test, necessitates the existence or use of a single or cutting score as a specified performance standard. However, much of the literature subsumed under the heading of measurement does, in fact, postulate the existence of a single cutting score. Since this inconsistency in terminology can lead to confusion, we prefer to reserve the term mastery for a with a single fixed mastery cutting score (see Harris, 1974). Hively (1974) and Millman (1974), among others, suggest using the descriptor test rather than criterion-referenced test. They note that the word criterion is ambiguous in some contexts, and they argue that the word domain provides a more direct specification of the entir, set of items or tasks under consideration. If one accepts these arguments, a mastery can be defined as a domain-referenced with a single cutting score. In this paper we develop and discuss an index of dependability for mastery tests. For reasons discussed later, we choose not to call our index a reliability coefficient, although many of the indices previously developed for mastery tests have been proposed as indices of reliability (see Brennan, 1974, for a review of the literature). For example, Livingston (1972) proposed a reliability coefficient based upon the squared deviations of scores from the cutting score; Swaminathan, Hambleton, and Algina (1974) proposed the use of Cohen's (1960) coefficient K; Marshall and Haertel (1975) suggest using a mean split-half coefficient of agreement; and Carver (1970) suggests two other coefficients.