Some further aspects of ecological regression analysis

✍ Scribed by Jan-Bernd Lohmöller; Jürgen W. Falter

• Book ID: 104652516
• Publisher: Springer Netherlands
• Year: 1986
• Tongue: English
• Weight: 782 KB
• Volume: 20
• Category: Article
• ISSN: 0033-5177
• DOI: 10.1007/bf00177966

✦ Synopsis

This paper is intended to serve three purposes. In the first part we introduce a predecessor to ecological regression analysis, which was first presented to the public as early as 1932 by the German statistician F. Bernstein. We will compare it with Leo Goodman's far better-known 1953 model of ecological inference. In the second part we discuss some applications of ecological regression analysis to Weimar election data with special emphasis on the NSDAP vote. Using available external evidence we are able to demonstrate that some variables such as religion and urbanization yield more plausible results than, for example, other variables such as sex or age. In part three we describe the statistical reasoning which helps to explain these differences and specifies some conditions under which we may reasonably expect "good", i.e. plausible inference from the aggregate to the individual level by means of ecological regression analysis.