Reliability: α vs. ω

Cronbach's α is not always the right reliability value — ω is often more honest.

What is this about?

How reliably does a questionnaire measure? α is the best-known answer, but holds only if all questions are related to the characteristic equally strongly. If the test additionally contains sub-topics, α can underestimate the reliability — or fake a unity that does not exist. ω decomposes the variance of the sum score more precisely: ω_total counts everything shared, ω_hierarchical only the part that goes back to the actual main construct. The tool shows the three values side by side and lets you change loadings and sub-topics — as with the 8-item cognition test with two factors but a single α.

How to proceed

1. Scenario & items — start with a preset (tau-equivalent, congeneric, multidimensional) and set the number of items, the loadings on the main construct (general factor) and optional sub-topics (group factors); strength 0 means unidimensional.
2. The three coefficients (①) — α, ω_total and ω_hierarchical compared.
3. Variance decomposition (②) — what the sum score is made of.
4. Split-half (③) — α as the mean of all test splits; plus the concepts.

What to watch for

With unequal loadings α underestimates the reliability because it weights every item equally.
With multidimensionality α deceives: α and ω_total also count the sub-topics as shared variance — only ω_hierarchical shows how much goes back to the main construct.
For 0/1 items KR-20 is nothing other than Cronbach's α.

Things to try

Choose the preset "multidimensional" and increase the strength of the group factors: α and ω_total stay high, ω_hierarchical falls — a single value would have faked unity here.

What's next?

Whether a test measures the same thing in different groups at all is what Measurement Invariance examines.

📋 What This Is About

Cronbach's α is the most-reported reliability coefficient — and the most frequently misused. It requires tau-equivalence (equal loadings) and unidimensionality. If these are violated, α can underestimate reliability (unequal loadings) or feign unidimensionality (multidimensionality) — exactly as with the 8-item cognitive test (V1–V4 verbal, S1–S4 spatial): two factors, but a single α. McDonald's ω is the more robust alternative — in two variants: ω_total and ω_hierarchical. This tool shows when each number misleads. Prerequisite: the measurement models (parallel/tau/congeneric).

← CTT Foundations: reliability as a proportion of variance

①The Three Coefficients

Cronbach's α

—

requires tau-equiv. & 1 dim.

ω_total

—

total shared variance

ω_hierarchical

—

general factor only

—

②Variance Decomposition of the Sum Score

What Does the Total Score's Variance Consist of? — and Where Do α, ω_t, ω_h Sit

③Split-Half — α ≈ Mean of All Test Halvings

Distribution of All Possible Split-Half Reliabilities (Spearman-Brown Corrected)

Concepts

When α Underestimates

With unequal loadings (congeneric), α — as an unweighted sum score — treats all items the same and wastes information: α < ω_total. α is then a lower bound on the true reliability. Solution: report ω_total.

When α Deceives

With multidimensionality (strong group factors), α and ω_total also count the group variance as "reliable." A high α then suggests a homogeneity that doesn't exist. ω_hierarchical shows how much really goes to the single general factor — often alarmingly little.

α = Mean of the Split-Halves

Split-half divides the test into two halves and correlates them (Spearman-Brown corrected). The result depends on how you split. Cronbach's α is exactly the mean over all possible halvings — panel ③ makes this visible.

KR-20 & Practice

For dichotomous (0/1) items, KR-20 is nothing other than Cronbach's α. The methodological literature (Revelle, McDonald, Flora) recommends reporting ω by default and, when unidimensionality is in doubt, additionally ω_h.

The Measurement Chain

This tool stands at the end of the chain: CTT Foundations (what is reliability) → Factor Analysis (estimate loadings from data) → Measurement Models (which assumption: parallel/tau/congeneric) → here (which coefficient is correct). The loadings ω computes with come directly from a factor model.
→ CTT Foundations → Factor Analysis → Measurement Models

Reliability: Cronbach's α vs. McDonald's ω