Description
Explain a two-variable relationship by recomputing it within categories of a third variable and reading the partial relationships as replication, explanation, interpretation or specification.
The elaboration model is a logic for multivariate analysis of survey data that predates and underlies today's stratified analysis, confounding and mediation checks and moderation. Starting from an observed zero-order relationship between two variables, the analyst introduces a test variable and recomputes the relationship within each of its categories, producing partial tables. The outcome is classified by two things: whether the test variable is antecedent to both variables or intervenes between them, and how the partials compare with the original. If the relationship holds in all partials it is replicated; if it vanishes with an antecedent variable it is explained away as spurious; if it vanishes with an intervening variable it is interpreted through a mechanism; if it holds in some partials and not others it is specified. Suppressor and distorter variables can hide or reverse a relationship. Ex post facto hypotheses must be tested on new data.
Attributes
- Discipline
- Statistics & Modelling
- Loss type
- Renamed
- Stage
- Evaluation, Post-launch
- Data
- Quantitative
- Effort
- Days
Steps
- Establish the zero-order relationship between the independent and dependent variable in the whole sample.
- Choose a test variable and justify whether it is antecedent to both or intervening between them.
- Split the sample by the test variable and recompute the relationship within each subgroup (partial tables).
- Classify: same in all partials is replication; reduced or vanished with an antecedent variable is explanation (spurious); reduced or vanished with an intervening variable is interpretation (mechanism).
- Classify as specification when the relationship is the same or stronger in one partial and weaker in another, and state the conditions.
- Check for suppressor or distorter variables that hide or reverse the relationship.
- Test any after-the-fact hypotheses with further data rather than resting on the observed tables.
Sources
- The Practice of Social Research (Babbie, 2013)
Also known as
- Lazarsfeld method
- Interpretation method
- Columbia school elaboration
- Partial-table analysis
- Sociological diagnostics
Categories
- Analytics & Behaviour
- Statistics & Modelling