Skip to content

Folders and files

NameName
Last commit message
Last commit date

Latest commit

 

History

37 Commits
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 

Repository files navigation

Research Edge Series

Rigorous Research Methodology for Social Scientists and Market Researchers

By Vinay Thakur · #vtmade

About the Series

The Research Edge Series provides evidence-based guidance for researchers who want to move beyond casual data collection to rigorous research practice. Each installment focuses on fundamental methodological challenges and their solutions, helping you avoid common pitfalls that destroy research validity.

Core Principles

Theoretical Rigor: Every recommendation is grounded in established research methodology • Practical Application: Complex concepts explained through real-world examples
Error Prevention: Focus on avoiding systematic biases that compound across studies • Actionable Insights: Move from "interesting findings" to strategic decisions

Article Index

Understanding Post-Hoc Rationalization in Consumer Research

Discover why consumers do not have conscious access to their decision-making processes and how post-hoc rationalization creates misleading research insights. Learn why direct "why" questions fail and how expert researchers use perception association methods to uncover real decision drivers.

Key Topics: • Post-hoc rationalization and System 1 vs System 2 thinking • Why direct questions about decision reasons fail • Perception association research techniques • Business impact of misunderstanding decision drivers

Available Formats: 📄 Article | 📁 PDF

Moving Beyond Casual Surveys to Rigorous Research Design

Learn why most research fails before data collection even begins. Discover the fundamental distinction between reflective and formative constructs, understand how measurement model misspecification can inflate your findings by 400%, and see how to fix the most common survey design errors.

Key Topics: • The consumer-as-researcher fallacy • Reflective vs. formative measurement models • Systematic bias in causal attribution • Step-by-step measurement design

Available Formats: 📄 Article | 📁 PDF

Why Statistical Precision Trumps Intuitive Mathematics

Contemporary business research suffers from a fundamental misunderstanding of statistical sampling theory, leading to systematic over-sampling and resource misallocation. Discover why population size bears minimal relationship to required sample sizes and how precision requirements should drive sampling decisions.

Key Topics: • The precision framework: Astronomical vs. engineering measurement standards • Mathematical reality vs. intuitive logic in population sampling • International research standards from WHO and UNICEF • Decision frameworks for optimal sample size determination

Available Formats: 📄 Article | 📁 PDF

Experimental Validation Using Real NYC Taxi Data

This supplementary article provides real-world experimental validation of the Sample Size Paradox using 1 million NYC taxi transactions. Demonstrates empirically how sample size affects estimate precision and validates the Central Limit Theorem through systematic sampling simulation.

Key Findings: • Dramatic accuracy improvements from n=30 to n=300 (6 percentage points precision gain) • Diminishing returns beyond n=500 samples • Statistical theory validated with 98% accuracy across all sample sizes • Visual demonstration of sampling variability through heatmap analysis

Supporting Materials: 🎯 Sampling Heatmap | 💻 Python Code | 📊 Data

Why Copying Successful Companies Usually Fails (And How Rigorous Research Fixes This)

Understand why most business advice misleads us by confusing correlation with causation. Learn the difference between anecdotal pattern-matching and rigorous causal analysis, including reverse causality, confounding variables, and how researchers use regression analysis and propensity score matching to establish probable cause.

Key Topics: • The halo effect and survivorship bias in business research • Three types of false causation: reverse causality, confounders, and pure correlation • Regression analysis and statistical controls explained • Propensity score matching: finding statistical twins • How to evaluate business advice critically

Available Formats: 📄 Article | 📁 PDF

Why Most AI Tools Flatten Qualitative Data, and What Rigorous Analysis Actually Requires

Understand why most AI-powered qualitative analysis produces summaries that look useful but fall apart under scrutiny. Learn what rigorous qualitative analysis actually involves — from coding and theme construction to pattern detection and cross-case analysis — and how to use AI as a structured coding layer that preserves traceability, confidence levels, and outliers.

Key Topics: • Why AI summaries of qualitative data are analytically useless • Grounded Theory and Thematic Analysis fundamentals • The Code → Theme → Pattern → Cross-case sequence • Six pattern lenses: similarity, difference, frequency, sequence, co-occurrence, cause • Structured analytical output vs. summary generation

Available Formats: 📄 Article

Claude Skill: A companion Claude Skill for structured qualitative analysis is available in the claude-skills directory.

Why Aggregate Quant Work Breaks LLMs, and the Workflow That Puts the Rigour Back In

Aggregate quantitative work is exactly where LLMs fail most often — and where teams are most eager to apply them. This article introduces a Prep → Validation → Analysis framework built around a flat-data contract that lets the LLM interpret while Python computes. Three prep paths handle every starting point: a crosstab from the agency, raw data destined for multivariate analysis, or raw data that needs cross-cut tables produced from scratch.

Key Topics: • Why raw respondent files and layout-heavy crosstabs both confuse LLMs • The flat-data shape that makes the rest of the workflow possible • Three prep paths: flatten / block extract / generate • Light-touch validation embedded in code, not workflow • Where the LLM earns its keep — code, interpretation, drafting findings

Available Formats: 📄 Article

Companion artefacts: flatten-crosstab skill · extract-crosstabs skill · tidy-data-analysis skill · Flat-format DP delivery spec

Why your best questions quietly return the wrong answer — and how to design around it

A survey is not a form. A form collects answers. An instrument is built to uncover, test, and decide. This article is about one specific way instruments break silently — asking respondents to perform a reasoning task they are cognitively incapable of performing — and the design discipline that fixes it. Covers attribute substitution, satisficing, the CASM four-stage response model, construct validity, and cognitive interviewing as pre-field detection.

Key Topics: • Attribute substitution and satisficing — the two invisible failure modes • The CASM four-stage model (comprehension → retrieval → judgment → response) • Why "just ask them to think harder" doesn't work • The decomposition principle: measure the answerable, infer the rest • Construct validity — Cronbach/Meehl and Campbell/Fiske frameworks applied to survey design • Cognitive interviewing as pre-field quality control

Available Formats: 📄 Article | 📁 PDF


Tools & Interactive Resources

Correspondence Analysis Map

A browser-based interactive tool for correspondence analysis. Paste or upload any grid data — brands × attributes, segments × motivations, markets × features — and generate a perceptual biplot. Handles the big-brand effect through built-in standardisation, supports draggable labels for cluttered maps, and exports publication-ready PNG, SVG, and the underlying coordinate scores as CSV.

Available on the live site — no installation needed, runs entirely in the browser. A companion Claude skill (correspondence-analysis.zip) is in the claude-skills directory.

A structured qualitative analysis pipeline for Claude — open coding, theme construction, six-lens pattern detection, full source traceability. See #005 for the companion article.

GenAI on Quantitative Survey Data — workflow skills

Three composable Claude Code skills plus a delivery specification, all designed to compose end-to-end with the flat-data format as the contract between them. Companion to #006.

Artefact What it does Where it sits
flatten-crosstab.zip Flattens agency crosstabs (XLSX/XLS/CSV) into the flat format. Inspect → confirm → flatten with a four-check validation report. Path A — crosstab in hand
Flat-format delivery spec DP instruction set that defines the flat output shape semantically — required columns, row-type taxonomy, value conventions. Worked examples and edge-case FAQ. Path A — flat format produced at source
extract-crosstabs.zip Generates flat-format crosstabs directly from raw respondent data (.sav/.csv/.xlsx). All five question types, weighting, conditional bases, custom NETs, optional significance testing. Path C — only raw data, need the tables
tidy-data-analysis.zip Picks up after prep — interactive analytical moves, evidence-pinned findings, four sanity checks per finding, resumable sessions. Analysis stage — flat file → defensible findings

Repository Structure

research-edge-series/
├── articles/           # Series articles in markdown format
├── assets/            # PDFs, images, diagrams, specifications, supporting materials
├── claude-skills/     # Claude Skills for research methodology
└── README.md          # This file

How to Use This Repository

  1. Browse Articles: Each article is self-contained and addresses a specific methodological challenge
  2. Follow the Series: Articles build on each other but can be read independently
  3. Apply the Frameworks: Use the decision trees and checklists in your own research
  4. Share and Contribute: Reference these materials in your work and suggest improvements

Citation

When referencing articles from this series, please use:

Thakur, V. (2026). [Article Title]. Research Edge Series #[Number]. 
Retrieved from https://github.com/vtmade/research-edge-series

Contact

For questions, feedback, or collaboration opportunities, reach out to Vinay Thakur at vpst18@gmail.com.


License

All materials are provided under the MIT License.

Disclaimer

This project is created by Vinay Thakur in a personal capacity for educational and knowledge-sharing purposes only. It does not represent, reflect, or endorse the views of any employer, organisation, or affiliated entity. The content is not intended for commercial use. You are free to use, adapt, and reference this work at your own discretion — please assess its suitability for your context independently.

Citation: If you use or reference this work, please cite:

Thakur, V. (2026). Research Edge Series. https://github.com/vtmade/research-edge-series

Research Edge Series · #vtmade

About

The Research Edge is an open-source platform I write and maintain for people who want to get better at research, whether you're doing the work, commissioning it, or making decisions from it. I publish articles on methodology, build browser tools that put proven methods into practice, and release Claude skills that researchers and managers can use

Topics

Resources

Code of conduct

Stars

Watchers

Forks

Contributors

Languages