From Pre-Survey to Valid Instruments in R

Building Post–Flash Flood Questionnaires with R

Flash floods are not merely hydrological events. They represent a collapse of social systems: homes destroyed, livelihoods lost, and security eroded. Yet many post-disaster studies fail to capture this reality—not due to lack of data, but because of weak instruments from the outset.

This article, prepared for Notivra Insight, provides a comprehensive guide to pre-survey design and validation of new questionnaires, using R as the primary analytical tool.

---

Why Pre-Surveys Are Non-Negotiable

A pre-survey is not administrative ritual. It is an epistemic filter. Its purposes are limited and precise:

1. To test whether items are understood as intended
2. To verify that constructs have a defensible structure
3. To eliminate flawed items before the main survey

Skipping this step undermines the entire research process.

---

Common Failures in Post-Disaster Research

Recurring methodological errors include:

• Proceeding directly to the main survey
• Treating Cronbach’s alpha as sufficient validation
• Ignoring factor analysis
• Retaining items for sentimental reasons

These are not minor flaws—they are design failures.

---

What Does “Validity” Really Mean?

There is no such thing as a “correct” questionnaire, only:

• Validity: does it measure the intended construct?
• Reliability: does it measure consistently?

For newly developed post-flood instruments, construct validity is critical and must be tested through Exploratory Factor Analysis (EFA).

---

Why Use R?

R offers clear advantages:

• Full transparency
• Reproducibility
• Acceptance in peer-reviewed journals
• Seamless transition to advanced models

R forces structural thinking, not button-clicking.

---

A Proper Pre-Survey Workflow

1. Theory-driven item construction
2. Pre-survey (30–100 respondents)
3. Data cleaning
4. KMO and Bartlett’s tests
5. Exploratory Factor Analysis
6. Item elimination
7. Reliability testing per construct

The outcome is a lean, valid, and defensible instrument.

---

Why EFA Cannot Be Skipped

Flash flood impacts are context-specific. Construct structures cannot be assumed. EFA tests whether respondents interpret items as researchers intend.

Without EFA, research relies on hope rather than evidence.

---

Essential Pre-Survey Outputs

A legitimate pre-survey must report:

• KMO and Bartlett statistics
• Final factor structure
• Factor loading tables
• Cronbach’s alpha per construct
• Justification for discarded items

This is the foundation of credible research.

---

Closing Note

Weak instruments lead to misleading conclusions and flawed policy.
A rigorous pre-survey is unforgiving—but scientific integrity demands it.

At Notivra Insight, we believe strong data are born from instruments willing to be tested, broken, and refined.