Understanding the differences and when to use each approach
For measurements that produce continuous numeric values
For measurements that result in discrete classifications
Typical Setup:
Data Collection:
Range Method:
ANOVA Method:
| Metric | Formula | Excellent | Marginal | Unacceptable |
|---|---|---|---|---|
| %GRR | (GRR / Total Variation) × 100 | < 10% | 10-30% | > 30% |
| NDC | 1.41 × (σpart / σGRR) | ≥ 5 | 3-4 | < 3 |
| P/T Ratio | Precision / Tolerance | < 0.1 | 0.1-0.3 | > 0.3 |
Typical Setup:
Data Collection:
Agreement Analysis:
Statistical Measures:
| Metric | Description | Excellent | Marginal | Unacceptable |
|---|---|---|---|---|
| Overall Agreement | % of classifications matching reference | ≥ 90% | 80-90% | < 80% |
| Kappa Value | Agreement beyond chance | ≥ 0.75 | 0.40-0.75 | < 0.40 |
| Within Operator | Operator consistency | ≥ 95% | 85-95% | < 85% |
When you might convert:
Process:
When you might convert:
Process:
| Application | Measurement Type | GRR Approach | Key Considerations |
|---|---|---|---|
| Machined part dimensions | Caliper measurements | Variable | Continuous values, tight tolerances |
| Thread inspection | Go/No-Go thread gage | Attribute | Pass/fail decision, functional requirement |
| Surface finish | Profilometer reading | Variable | Numeric Ra values, process control |
| Visual defect inspection | Operator judgment | Attribute | Subjective, multiple defect types |
| Torque testing | Torque wrench reading | Variable | Continuous values, safety critical |
| Color matching | Visual comparison | Attribute | Subjective, categorical decisions |