Cross-volume method chaining (DSRP to ethnography to DAG to SEM to GMM)

Description

A deliberate sequence in which each method's output becomes the next method's input, moving from systems framing to ethnographic mechanism to causal graph, measurement model and segmentation.

Method chaining makes the hand-offs between methods explicit and testable. DSRP (Distinctions, Systems, Relationships, Perspectives) frames the problem space and surfaces candidate constructs. Ethnography grounds those constructs in observed practice and mechanisms. The mechanisms are formalised as a directed acyclic graph that states which variables cause which, and which are confounders. The DAG's latent constructs are then measured with a validated survey and estimated in a structural equation model, and the resulting factor scores or behavioural features feed a Gaussian mixture model that segments users probabilistically. The rigour lies in traceability: every SEM path should trace back to an ethnographic observation, every segment should be interpretable through the SEM constructs, and each link carries its own validity check rather than inheriting confidence from the previous stage.

Attributes

Loss type
Community-proposed
Data
Mixed
Effort
Months

Steps

  1. Map the problem with DSRP, listing distinctions, part-whole systems, relationships and stakeholder perspectives, and extract 5-12 candidate constructs.
  2. Run ethnographic fieldwork (typically 10-20 participants) to confirm, discard or refine each construct and to describe causal mechanisms in participants' terms.
  3. Translate mechanisms into a DAG, mark confounders, mediators and colliders, and check it with domain experts before measurement.
  4. Develop or adopt multi-item scales (at least 3 items per construct) and collect survey data large enough for SEM (commonly N of 200 or more).
  5. Fit a CFA then SEM, accepting the measurement model only at acceptable fit (e.g. CFI of 0.95 or above, RMSEA of 0.06 or below, SRMR of 0.08 or below).
  6. Fit GMMs on factor scores or behavioural features, choose the number of components by BIC and interpretability, and require each segment to be at least about 5% of the sample.
  7. Write a traceability matrix linking each segment and path back to field evidence, and flag links supported by only one stage.

Sources

  • Forward-Thinking Playbook (Gopinath)
  • Nash Equilibrium Playbook (Gopinath)
  • Behavioral Drift Playbook (Gopinath)
  • Cabrera, D. & Cabrera, L. (2015) Systems Thinking Made Simple
  • Pearl, J., Glymour, M. & Jewell, N. P. (2016) Causal Inference in Statistics: A Primer
  • Kline, R. B. (2023) Principles and Practice of Structural Equation Modeling, 5th ed.
  • The Inquiry Series, Ch 60 — The Forward-Thinking Research Loop — Predict → Diagnose → Prove → Forecast → Prescribe → Validate

Also known as

  • Method chaining
  • Mixed-methods pipeline chaining

Categories

  • Community-Proposed Frameworks