Different kinds of evidence#
| Evidence | Question it addresses | Scope limit |
|---|---|---|
| Analytical reference checks | Does the calculation agree with a known reference case? | Agreement in the defined case does not establish accuracy for every room. |
| Implementation and integration checks | Are inputs, calculations, saved results and interfaces handled consistently? | A successful workflow or repeatable output is not a measured accuracy result. |
| Independent measured comparison | How do specified predictions compare with measurements? | Applies to the tested version, configuration, geometry, metrics and bands. |
Verification also includes analytical reference checks and implementation/integration checks. Each addresses a different question. Numerical support and repeatability do not by themselves establish agreement with measurements.
What a published evidence case must identify#
- Date, tested software version and relevant configuration.
- Geometry, materials, source and receiver inputs, and the independent reference.
- The specific metric and frequency bands assessed.
- Acceptance criteria, observed results, coverage and outliers.
- Known limitations and remaining host/scientific checks.
A published definition is not proof of standards conformance. Project target checks and model predictions do not certify the installed space. For an evaluation decision, review evidence for the particular workflow and output you intend to use.
Calculation assumptions and limitations#
Room Analysis — statistical STI potential#
Room Analysis STI potential is reverberation-only screening with ideal-quiet assumptions; it does not include direct-field or electroacoustic corrections. It uses the 4 kHz reverberation value as a proxy for 8 kHz. It must not be presented as a measured or fully validated STI result.
Acoustic Ray Analysis — receiver-based STI estimate#
Ray speech estimates depend on the receiver response and saved background context. Not specified background, explicitly ideal quiet and entered noise are distinct scenarios. An unspecified noise context must not be silently described as zero noise.
The Acoustic Ray Analysis speech estimate does not include auditory-masking or hearing-threshold corrections.
For the broader modelling and output scope, see Model and output limitations.