Example 4 — Pressure transmitter, model provided directly

Exercises the Model (transfer function) wizard path — for sensors that arrive with a calibration certificate stating the calibration model directly, with no raw reference/measured data points to re-fit.

4.0 Setup

Asset (Assets → New Asset):

One channel:

The accuracy spec is what the model-provided-directly conformity check compares against — see below.

4.1 Wizard — Step 1 (General info)

4.2 Wizard — Step 2 (Input data)

Step 2 opens with a method picker. Choosing Model (transfer function) swaps the rest of the step from raw-data entry to a model-entry form.

The certificate states: reference ≈ 1.0025 × measured − 0.42 kPa, valid 0–1000 kPa, reference standard uncertainty ±0.5 kPa (k=2).

The equation preview reads: $f(x) = 1.0025x - 0.42\ \text{(kPa)}$.

4.3 Wizard — Step 3 (Review) and what gets saved

Even with no raw data, Step 3 collects the same uncertainty budget inputs a raw-data calibration collects:

Behind the scenes, the wizard evaluates two synthetic zero-error points at the declared valid range (0 and 1000 kPa) through the same analysis pipeline raw data uses, with the fit skipped — this combines the Type B budget above (here, just the reference standard term) and applies the decision rule with the model trusted exactly as declared. On save:

• poly_order = 1, poly_coefficients = [1.0025, -0.42], range_min = 0, range_max = 1000, valid_range_min = 0, valid_range_max = 1000.
• uncertainty_budget has one Type B row (reference_standard, standard uncertainty 0.5 / 2 = 0.25 kPa) — the Type A (fit_residuals) row is present but exactly 0, since the model is trusted as declared.
• combined_uncertainty ≈ 0.25 kPa, expanded_uncertainty ≈ 0.49 kPa (k≈1.96 at 95% confidence, normal distribution).
• r_squared, rmse, max_error, and non_linearity are all null — there's no real measured data to derive them from (see Alternative calibration data-entry modes).
• decision_rule = "guard_band_w_uncertainty" and a real conformity_statement: the acceptance zone is shrunk inward by the expanded uncertainty (0.5 - 0.49 ≈ 0.01 kPa of margin left), so this record CONFORMS — barely, illustrating exactly what guard-banding is for.

Expected results