A Two-Decision Model for Responses to Likert-Type Items Anne Thissen-Roe David Thissen
Out line • Extreme Response Measure • Models for Extreme Responding • Multidimensional Nominal Categories Model • Sequential Decisions Model for Extreme Response • Two Decision Model • Empirical Study • Discussion
Extreme Response Measures • Age: positive correlation (some nations) • Women choose more extreme than men? • Educational level: negative correlation • More anxiety more extreme • Low social interest (altruism & empathy)
Measurement of extreme response • Number of options, ambiguity, emotional arousal and speededness. • Content independent • Score inflation • Strategies: • ipsative rescaling, treat it as within subject SD. • Creation of an extreme scale • Model the response as additional psy variables.
Models for extreme responding • Integration for construct measurement and extreme responding • The choice of a parametric model • The number of latent variable • The parametric to describe the extreme response • Which response are assumed to equal across items • Identification constraints
Models for extreme responding • Johnson (2003) • Proportional thresholds model (PTM)(extreme shift the threshold) • Javaras and Ripley (2007) • Unfolding model(extreme shift the threshold) • Bolt and Johnson (2009) • Multidimensional nominal response model (MNRM)
A multidimensional nominal categories • precludes modeling items with categories in different orders along different axes
Discussion • 1. No strict experiment was conducted to confirm this proposed two-stage process, so that we can have many versions of stories to assume the model other than that one in the paper. The model lacks empirical evidence. Note that model fit is never ever a piece of evidence based one-stage data they used! • 2. A person may not really exhibit his or her 'extremeness' but too less number of categories enforced him or her to choose one out of them. • 3. Indulging ourselves in the same fantasy, unfolding model can be easily applied to the same story.