Interpretation of results
How to read the output of an evaluation — relative response and its margin, the three verdict states, withheld verdicts, marginal results, and what an overall verdict does and does not assert.
Last reviewed
The output of an evaluation is a set of computed quantities and a set of verdicts. Reading it correctly requires knowing which quantity carries the decision and what each verdict state asserts.
Computed quantities
| Quantity | Definition | Role |
|---|---|---|
| Mean value, G | Mean of the repeated readings, less the background mean | Intermediate |
| Response, R | G divided by the reference value | Intermediate; carries the instrument’s calibration factor |
| Relative response, Rr | R divided by the response at the reference point | The acceptance quantity |
| Uncertainty, ± u | Type A standard uncertainty, propagated | Reported; not applied to the verdict |
Response and relative response are frequently confused. Response is not the acceptance quantity. It contains the instrument’s overall calibration factor, so an instrument that reads uniformly 20 % low has a response of 0,8 everywhere and a relative response of 1,00 everywhere. It passes.
That is correct behaviour. These tests evaluate whether the response varies — with energy, with angle, with dose rate. A uniform offset is a calibration matter and is outside what these tests assess.
Verdict states
Three states, and the third is not a variant of the second.
Pass
Every point in the test is inside its acceptance limit.
Fail
At least one point is outside its limit. There is no averaging and no majority rule: one point outside fails the test.
Incomplete
The test could not be judged. This state exists so that an unjudged test and a passed test are distinguishable in a report. It occurs when:
- required data is missing — for example H₀ for statistical fluctuation, or the 60 s indication for response time;
- a checklist test has only one of its two required observations answered;
- fewer than three points are available where the acceptance rule requires three.
Incomplete is not a soft fail. It asserts nothing about the instrument. Treating it as a pass in a report is a misstatement; treating it as a fail attributes a defect that was not demonstrated.
Margin
The relevant reading is not the verdict alone but the distance from the value to the limit.
A relative response of 1,21 against an upper limit of 1,22 and one of 1,02 against the same limit are both recorded as a pass. They are not the same result. The first has a margin of 0,01 and a Type A uncertainty that is normally larger than that; the second has a margin of 0,20.
Compare the margin against the reported uncertainty. Where the margin is smaller, the correct statement in a report is that conformity could not be demonstrated at that point, not that the point passed. The tool does not make this judgement, for the reasons set out in the uncertainty guide.
A marginal result is also the case most likely to change between two evaluations of the same instrument, which is a practical reason to record the margin rather than the verdict alone.
Overall verdict
The overall verdict combines the individual tests:
- Fail if any test failed;
- otherwise Incomplete if any required test is incomplete;
- otherwise Pass.
Treatment of optional tests
An optional test affects the overall verdict only if it was performed. A test left empty is neither a pass nor a failure and does not enter the combination. A performed optional test that fails does make the overall verdict a failure.
The consequence is that an overall pass does not assert that the instrument satisfies every characteristic in the standard. It asserts that it satisfied the ones evaluated. The list of tests performed is part of the result and a report that omits it is not interpretable.
Charts
Each test with a numeric acceptance band presents the points against the band. The band is shaded; points outside it are marked.
For statistical fluctuation the chart shows CV divided by its limit rather than CV itself, because the limit steps with dose rate. Plotting CV directly against a single line would misrepresent the acceptance condition. On that chart the shaded region is the c₁ condition, and the region between c₁ and c₂ is where up to two non-adjacent points are permitted.
Excel export
The export contains one sheet per test with the entered readings, the computed quantities, the per-point verdicts and the overall verdict, together with the disclaimer.
The exported workbook is the record. Session history is held in the browser’s local storage and is removed when browser data is cleared; it cannot be recovered, because it is not held remotely. Export anything that must be retained.