Attribute Agreement Analysis
Indice
Variable dados versus atributo dados
The distinction determines which estudo you can run, and it is decided by what the inspection actually records, not by what is being inspected.
| Aspect | Variable dados | Attribute dados |
|---|---|---|
| O que is recorded | A measured value on a continuous scale, such as 25.012 mm | A category, such as pass or fail, or a grade such as A, B or C |
| Applicable estudo | R&R do sistema de medição, viés, linearidade, estabilidade | Attribute concordância análise |
| Headline resultado | Percent GRR and ndc | Percent concordância and kappa |
A useful test: if you cannot meaningfully subtract one resultado from another, the dados is atributo dados. Pass minus fail is not a quantity, so no variância can be decomposed and the entire R&R do sistema de medição machinery does not apply.
Por que you cannot just run a R&R do sistema de medição
R&R do sistema de medição works by splitting the total observed variância into peça variação, repetibilidade and reprodutibilidade. Every one of those terms is a variância, and a variância requires a numeric scale with meaningful distances between values. Category labels have no such scale.
People sometimes code pass as 1 and fail as 0 and feed that into an R and R calculation. Do not do this. The arithmetic will produce numbers, and those numbers will be meaningless, because the distance between the codes is an artifact of the coding rather than a propriedade of the medição. An auditor who understands the método will recognize it immediately.
The three comparisons
A complete atributo estudo has each avaliador inspect the same set of peças more than once, in a randomized order, without knowing the previous resultado. That design supports three distinct perguntas.
Within avaliador
Does one avaliador reach the same verdict when shown the same peça again? This is the atributo equivalent of repetibilidade. Poor within-avaliador concordância means the criteria are ambiguous or the defect is near the limit of what can be seen.
Between avaliadores
Do different avaliadores reach the same verdict on the same peça? This is the atributo equivalent of reprodutibilidade. Poor between-avaliador concordância usually points at inconsistent training or an unwritten padrão that each person has interpreted differently.
Versus the padrão
Do the avaliadores reach the correct verdict, judged against an independently established referência decision for each peça? This is the one that matters most, and the one that requires the most preparation.
Kappa: concordância corrected for chance
Raw percent concordância overstates how good a processo is, because two people guessing randomly will still agree a large fraction of the time. With only two categories, pure guessing produces about 50 percent concordância. Cohen's kappa removes that baseline.
Cohen's kappa
κ = (Po − Pe) ÷ (1 − Pe)
where Po is the observed proportion of concordância and Pe is the proportion of concordância expected by chance alone, calculated from how often each avaliador uses each category.
The numerator is the concordância above chance actually achieved. The denominator is the concordância above chance that was available to achieve. Kappa is therefore the fraction of the achievable improvement over guessing that the avaliadores captured. Kappa of 1 is perfect concordância and kappa of 0 is no better than chance. Negative values are possible and indicate systematic disconcordância.
Worked example
Illustrative dados. Two avaliadores each classify the same 50 peças as pass or fail.
- They reach the same verdict on 45 of the 50 peças, so Po = 45 / 50 = 0.90.
- Both avaliadores call about 80 percent of peças pass and 20 percent fail, so the chance concordância is Pe = (0.80 x 0.80) + (0.20 x 0.20) = 0.64 + 0.04 = 0.68.
- kappa = (0.90 - 0.68) / (1 - 0.68) = 0.22 / 0.32 = 0.69.
Ninety percent concordância sounds strong. Kappa of 0.69 tells a different and more honest story: most of that concordância was available for free because the processo rejects few peças. This gap between raw concordância and kappa is exactly why the correction exists, and it grows wider as the categories become more unbalanced.
Interpreting kappa
| Kappa | Common interpretation |
|---|---|
| κ ≥ 0.90 | Good concordância. The inspection is generally considered acceptable on this criterion. |
| 0.70 ≤ κ < 0.90 | Marginal. Usually treated as needing improvement to the criteria, the training, or the inspection conditions. |
| κ < 0.70 | Unacceptable. The inspection is not reliably distinguishing conforming from non-conforming peças. |
Treat a low kappa as a processo finding rather than a people finding. It almost always traces back to criteria that were never written down precisely, boundary samples that do not exist, or lighting and fixturing that make the defect genuinely hard to see. Retraining without fixing those conditions rarely moves the number.
O que you can do in GaugeConnection today
Until that changes, the practical steps are:
- Run the estudo outside this tool against your own documented procedure, and retain the records where your quality system requires them.
- Before running it, write down the acceptance criteria and produce physical boundary samples. Most atributo estudos fail because the padrão was never made concrete, and no statistic will repair that.
- Onde an atributo check can be replaced by a medição, consider doing so. Variable dados gives you far more information per peça and supports the full faixa of MSA estudos.