Analysis of variance on ordinal variables: application to opinion research

✍ Scribed by Bernard, S. ;Derquenne, C. ;Oger, P.

• Publisher: John Wiley and Sons
• Year: 1997
• Tongue: English
• Weight: 119 KB
• Volume: 13
• Category: Article
• ISSN: 8755-0024
• DOI: 10.1002/(sici)1099-0747(199709/12)13:3/4<345::aid-asm328>3.0.co;2-y

✦ Synopsis

Medical research data are often modelled either by analysis of variance (continuous response variables) or by logistic regression (nominal or ordinal response variables). These methods allow one to eliminate the structural effects among variables which are candidates for an explanation and to measure these effects, other things being equal. Our department applies these methods to opinion surveys. The quality and validity of results are very important because strategic decisions may depend on them. In this paper, we focus on the modelling mechanism and the statistical tests. After, we apply our model on two examples: French nuclear policy and the Maastrich referendum. Lastly, we conclude on two future issues in the domain of logistic analysis.

1998 John Wiley & Sons, Ltd.

KEY WORDS polytomous models; logistic analysis; maximum likelihood; Berkson's method 1. ANALYSIS OF VARIANCE ON ORDINAL DATA: ESTIMATION METHODS Let ½ be a response variable (numeric), and X J , 2 , X N , p 'explanatory' variables. If the number of explanatory variables is, for instance, two, we often use an analysis of variance model: