To understand the relationship between variables of social and political nature, social scientists have a strong interest in regression-style specifications such as Poisson, logistic, multinomial, and more. Although these specifications are unnecessarily treated as distinct and particular, they are all part of the generalized linear models (GLMs), a class of regression models that share a common theoretical basis and structure. This unified framework facilitates the understanding and derivation of the properties of these models, as well as a more principled application of them to actual social science data. This entry introduces and explains the underlying structure of GLMs, demonstrates the theoretical basis that their various forms share, and illustrates their application with two running examples. A proper understanding of the GLM framework can increase a researcher’s flexibility with regard to new data types.
By: Jeff Gill & Michelle Torres | Edited by: Paul Atkinson, Sara Delamont, Alexandru Cernat, Joseph W. Sakshaug & Richard A. Williams Published: 2020 | Length: 10 | DOI: http://dx.doi.org/10.4135/9781526421036883329 |