MeasurementUncertaintyEvents¤

Measurement uncertainty per the GUM (ISO/IEC Guide 98-3): Type A from the scatter of repeated readings per window, Type B from sources you supply (calibration certificate, resolution, temperature), combined and expanded uncertainty, and the uncertainty budget.

ts_shape.events.quality.measurement_uncertainty · Quality guide · API reference

Example¤

import ts_shape
from ts_shape.events.quality.measurement_uncertainty import MeasurementUncertaintyEvents

df = ts_shape.make_timeseries(["cmm1:bore_diameter"], n_points=2000, freq="1min", baseline=25.0, noise=0.003)
sources = [
    {"source": "calibration", "half_width": 0.004, "distribution": "normal"},
    {"source": "resolution", "half_width": 0.0005, "distribution": "rectangular"},
]

mu = MeasurementUncertaintyEvents(df, "cmm1:bore_diameter")
result = mu.combined_uncertainty(window="8h", type_b_sources=sources, coverage_factor=2.0)
print(result[["start", "type_a", "combined_uncertainty", "expanded_uncertainty"]])
budget = mu.uncertainty_budget(window="8h", type_b_sources=sources)

Methods¤

Method	Returns
`type_a_uncertainty(window='8h')`	per window: `start`, `end`, `duration_seconds`, `mean`, `std`, `sample_count`, `type_a_uncertainty`
`type_b_uncertainty(sources)` (static)	per source: `source`, `distribution`, `half_width`, `divisor`, `type_b_uncertainty`
`combined_uncertainty(window='8h', type_b_sources=None, coverage_factor=2.0)`	per window: `type_a`, `type_b_combined`, `combined_uncertainty`, `expanded_uncertainty`, `coverage_factor`
`uncertainty_budget(window='8h', type_b_sources=None)`	per source: `source`, `uncertainty`, `contribution_pct`

Notes¤

Type A: u_A = std / sqrt(n) per window; windows with fewer than 2 samples are skipped.
Type B: u_B = half_width / divisor, with divisors rectangular √3 (default, also for unknown names), triangular √6, normal 2, u-shaped √2. Each source needs source and half_width; distribution is optional.
Combined: u_c = sqrt(u_A² + Σ u_Bi²); expanded: U = k · u_c with k = coverage_factor (2.0 ≈ 95.45 % coverage).
Budget: Type A is averaged over all windows; contribution_pct is each source's share of the total variance (u²), so the rows sum to 100.
part_uuid is deprecated and has no effect.