Strengths and limitations of ipsative measurement

Abstract

The relative merits of ipsative measurement for multi-scale psychometric instruments are discussed. Standard psychometric analyses are found to be inappropriate with small numbers of scales, or sets of scales with strong positive intercorrelations. However, for larger sets of scales ( N -30) with low average intercorrelations, ipsative data seems to provide robust statistical results in reliability analyses but not under factor analysis. It is argued that the potential of forced-choice formats to control for some of the response biases typical of normative scales means that they could have substantial advantages while data from score distributions are used to show that for higher numbers of scales, the vast majority of profiles are not so skewed that they would be likely to be distorted by a forced-choice response format. The implications for score interpretation are discussed.

Impact of situational framing and complexity on charismatic, ideological and pragmatic leaders: Investigation using a computer simulation

“…Third, the leader measure employed in the study was ipsative in nature and is therefore only appropriate for categorizing leaders into preferred types (Baron, 1996;Chan, 2003). More directly, total scores on the various leader types (e.g., a participant's degree of pragmatic leadership) could not be obtained.…”

How IRT can solve problems of ipsative data in forced-choice questionnaires.

“…Principal components analysis can be performed on ipsative scores; however, it may yield quite different results from the analysis of corresponding normative scores. Ipsativity produces artifactual bipolar components, where traits that would typically belong to different components in normative data are contrasted to each other Baron, 1996). Ipsativity, therefore, clearly compromises construct validity of forced-choice questionnaires.…”

“…For instance, Baron (1996) argued that coefficient alpha underestimates internal consistency reliabilities in MFC questionnaires measuring a large number of scales. On the contrary, Johnson, Wood and Blinkhorn (1988) suggested that reliabilities of ipsative tests overestimate the actual scale reliabilities.…”

“…Baron (1996) argued that with a large number of relatively independent scales, only a very low proportion of respondents will have most of their true scale scores all high or all low -and therefore only a few profiles will be badly distorted by the ipsative centering on the mean score. Also, several authors have shown that in carefully designed forcedchoice questionnaires with 30 or more measured traits, the ordering of people on each trait largely corresponds to their normative ordering (e.g., Baron, 1996;Karpatschof & Elkjaer, 2000), and therefore the standardizing of ipsative scores is appropriate and inter-individual comparisons can be performed meaningfully.Because the average correlation in Equation (1) approaches zero as the number of scales increases, with 30 scales the average correlation would be approximately 0.03, allowing for a wide range of both negative and positive correlations between scales (Baron, 1996). As a result, principal components analysis of ipsative scores may also be more interpretable, although it does not produce the same results as analysis of normative scores.…”

Space closing versus space opening for bilateral missing upper laterals – aesthetic judgments of laypeople: a web-based survey

“…In psychological research the ipsative style has been used to avoid the potential for central tendency bias with a Likert response format, as respondents avoid the most extreme categories (Baron, 1996, Christiansen et al, 2005. Other potential biases with Likert scales include acquiescence responding, where responders tend to agree with statements and social desirability responding where they try to appear more positive (and occasionally more negative) than they are.…”