CHAPTER 16 – THE ELABORATION MODEL
Elaboration model: a logical approach to understanding the relationship
between 2 variables thru the simultaneous introduction of a 3rd variable, usually
referred to as a control/test variable. Though developed primarily thru the
medium of contingency tables, it may be used with other statistical techniques.
The various outcomes of an elaboration analysis include replication,
specification, explanation, & interpretation.
It aims to elaborate on an empirical relationship among variables in order
to interpret that relationship
The central logic begins w/ an observed relationship between 2 variables
& the possibility that one variable may be causing the other.
Sometimes the analysis reveals the mechanisms thru which the causal
relationship occurs. Other times an elaboration analysis disproves the
existence of a causal relationship altogether.
Having observed an empirical relationship between 2 variables (level of
education & acceptance of induction), we seek to understand the nature of
that relationship thru the effect produced by introducing other variables
(such as having friends who were deferred). We do this by 1st dividing our
sample into subsets on the basis of the test variable (control variable).
Test variable: a variable that is held constant in an attempt to clarify
further the relationship between 2 other variables.
EX: Having discovered a relationship between education &
prejudice, we might hold gender constant by examining the
relationship between education & prejudice among men only &
then among women only. In this ex, gender is the test variable.
The relationship between the original 2 variables (acceptance of induction
& level of education) is then recomputed separately for each of the
subsamples. The tables produced in this manner are called the partial
tables, & the relationships found in the partial tables are called the partial
Partial relationships: the relationship between 2 variables when
examined in a subset of cases defined by a 3rd (test) variable.
EX: beginning w/ a zero-ordered relationship between income &
attitudes toward gender equality. We want to see whether the
relationship holds true among ♂ & ♀ (that is, controlling for
gender). The relationship found among ♂ & the relationship found
among ♀ would be the partial relationships, aka the partials.
The partial relationships are then compared w/ the initial relationship
discovered in the total sample, often referred to as the zero-order
relationship to indicate that no test variables have been controlled for.