Quality & Quantity 37: 337–361, 2003.
© 2003 Kluwer Academic Publishers. Printed in the Netherlands.
Inferential Causal Models: Integrating Quality &
ROBERT B. SMITH
Social Structural Research Inc., 3 Newport Rd., Suite 6, Cambridge, MA, U.S.A.
Abstract. In-depth data analysis plus statistical modeling can produce inferential causal models.
Their creation thus combines aspects of analysis by close inspection, that is, reason analysis and
cross-tabular analysis, with statistical analysis procedures, especially those that are special cases
of the generalized linear model (McCullagh and Nelder, 1989; Agresti, 1996; Lindsey, 1997). This
paper explores some of the roots of this combined method and suggests some new directions. An
exercise clariﬁes some limitations of classic reason analysis by showing how the cross tabulation of
variables with controls for test factors may produce better inferences. Then, given the cross tabulation
of several variables, by explicating Coleman effect parameters, logistic regressions, and Poisson
log-linear models, it shows how generalized linear models provide appropriate measures of effects
and tests of statistical signiﬁcance. Finally, to address a weakness of reason analysis, a case-control
design is proposed and an example is developed.
Key words: reason analysis, generalized linear models, case-control studies.
1. Close Inspection of Data
Reason analysis and cross-tabular analysis are both methods for the qualitative
analysis of surveys (Kadushin, 1968, 338). The reason analyst selects cases on the
basis of a category of a dependent variable, which is an action (such as going to
a psychiatrist) or a choice (such as intending to vote for a Democratic candidate).
Then, in an interview the reason analyst probes and classiﬁes a person’s subjective
reasons for performing that particular action or for making that choice. After a
sample of people have been interviewed and their reasons for each of their actions
or choices assessed, the analyst then aggregates their responses to form variables
and distributions for the whole sample of interviewees. Contrariwise, the analyst
of survey cross tabulations assesses coded interview data for a whole sample of
people – some of whom may have acted one way and others another way – and
then assesses the statistical relationships among the variables of interest.
2. Classic Reason Analysis
In his explications of the paradigm of reason analysis and of the meaning of cause,
Lazarsfeld begins with a hypothetical pure experiment (, 1955, 412); Table I
reproduces its design and data: