Abstract
Marginal modeling and causal analysis are often seen as two opposite and mutually exclusive approaches. This is understandable because marginal modeling is frequently defined in contrast to conditional analyses, and the outcomes of conditional analyses are typically interpreted in causal terms. However, this opposition of marginal and causal analyses is wrong or at least way too simple. For instance, most, if not all, marginal models discussed in the previous chapters might have been derived from causal theories. The tests on the validity of these marginal models could then be regarded as tests on the validity of these underlying causal theories. Further, marginalmodeling is often an indispensable part of approaches that are explicitly intended to investigate causal relationships, such as (quasi-)experimental designs and causal models in the form of structural equation models (SEMs). In this chapter, the focus will be on the role of marginal modeling in these last two approaches. Although this chapter is about ‘causal modeling,’ the term ‘causality’ itself will not be used very often (unless it leads to cumbersome and awkward phrases). Unfortunately, it is still common social science practice to use ‘causality’ in a rather weak and general sense, more pointing towards some assumed asymmetry in the relationships rather than necessarily implying proper causal connections. The disadvantages and dangers of this careless use of the term are first of all, that when drawing the final conclusions researchers tend to forget the caveats surrounding their causal analyses, evenwith research designs and research questions that hardly permit inferences about causal connections; and second, that researchers do not think hard enough before the start of their investigations about the appropriate research design for answering their causal questions. Therefore, even when the term causality is used below, the reader should keep these reservations in mind. For just a few treatments of causality with different emphases, see Rubin (1974), Cartwright (1989), Sobel (1995),McKim and Turner (1997), Sobel (2000), Pearl (2000), Snijders and Hagenaars (2001), Winship and Sobel (2004),Heckman (2005) (with discussion), and Morgan andWinship (2007), and the references cited in these publications.
This is a preview of subscription content, log in via an institution.
Buying options
Tax calculation will be finalised at checkout
Purchases are for personal use only
Learn about institutional subscriptionsPreview
Unable to display preview. Download preview PDF.
Author information
Authors and Affiliations
Corresponding author
Rights and permissions
Copyright information
© 2009 Springer-Verlag New York
About this chapter
Cite this chapter
Bergsma, W., Croon, M., Hagenaars, J.A. (2009). Causal Analyses: Structural Equation Models and (Quasi-)Experimental Designs. In: Marginal Models. Statistics for Social and Behavioral Sciences. Springer, New York, NY. https://doi.org/10.1007/b12532_5
Download citation
DOI: https://doi.org/10.1007/b12532_5
Published:
Publisher Name: Springer, New York, NY
Print ISBN: 978-0-387-09609-4
Online ISBN: 978-0-387-09610-0
eBook Packages: Mathematics and StatisticsMathematics and Statistics (R0)