Study guide

Propagation, Covariance, and Effective Degrees of Freedom

ASQ Certified Calibration Technician CCT

Track your progress as you read

Finished sections get a check mark.

1 SI Definitions and Realizations

The Seven SI Base Units and the 2019 Revision Realizing SI Units in the Laboratory

2 Derived Units, Prefixes, and Constants

Derived Units and Physical Calculations SI Prefixes and Compound-Unit Conversions Physical Constants and Their Applications

3 Measurement Standards and Selection

Types and Roles of Measurement Standards Selecting Standards for an Operational Hierarchy

4 Traceability and Standard Substitution

Traceability Evidence and Impact Review Evaluating and Authorizing Standard Substitution

5 Dimensional and Mechanical Metrology

Linear Displacement and Dimensional Instruments Geometry, Thermal Expansion, and Mass Force, Torque, Pressure, Density, and Viscosity

6 Temperature, Humidity, Volume, and Flow

Temperature Sensors and Thermometer Corrections Humidity, Volume, and Flow

7 Electrical, Optical, and Time Measurements

Voltage, Current, Resistance, and AC Measurement Time, Frequency, Light, Power, and Energy

8 Measurement Methods

Direct, Indirect, and Ratio Methods Transfer, Differential, and UUT Substitution Methods

9 Instrument Characteristics and Measurement Data

Sensitivity, Variability, and Precision Bias, Linearity, Stability, and Hysteresis Data Formats, Readability, and Suitability Resolution, Confidentiality, and Data Integrity

10 Specifications and Error Sources

Interpreting IM&TE Specification Terms Specification Calculations and Operating Qualifiers Instrument, Operator, and Circuit Errors Environmental, Geometric, and Interference Errors

11 Measurement Assurance

Check Standards and Statistical Process Control Proficiency Comparisons and Gage R&R

12 Procedures and Governing Requirements

Procedure Structure and Published Practice Sources Method Development, Deviations, and Handling Regulations, Contracts, and Requirement Precedence

13 Instrument Adjustment and Standardization

Calibration, Zero, and Span Adjustment Linearization, Nulling, and Digital Adjustment

14 Calibration Environment

Temperature and Humidity Control ESD, Dust, Vibration, and Pressure

15 Operations, Roles, and Scheduling

Receiving, As-Found Testing, and Calibration Flow Asset Identification, Tracking, and Release Laboratory Roles and Initial Scheduling Interval Review, Overdue Control, and Workloads

16 Method and Software Validation

Method Verification and Validation Studies Software, Firmware, and Spreadsheet Validation

17 Records and Formal Reporting

Controlled Records and Calibration Certificates Status Labels and Out-of-Tolerance Reporting

18 Uncertainty Concepts and Decision Risk

Error, Uncertainty, and Measurand Definition TAR, TUR, Tolerance Consumption, and Capability Conformity Decisions and False-Accept Risk Guard Bands and Conformity Reporting

19 Uncertainty Models and Budgets

Measurement Models and Uncertainty Contributions Distribution Selection and Budget Assembly

20 Type A and Type B Evaluation

Type A Statistics and Repeatability Outlier Review and Worked Type A Analysis Type B Distribution Models

21 Uncertainty Reporting and Applied Mathematics

Propagation, Covariance, and Effective Degrees of Freedom Coverage Factors and Expanded-Uncertainty Reporting Notation, Unit Conversions, Ratios, and Decibels Rounding, Algebra, Graphs, and Angular Conversions

22 Quality Systems and Tools

Quality-System Components, Planning, and Improvement The Seven Basic Quality Tools

23 Audits and Corrective Action

Audit Types, Planning, and Nonconformance Control Impact Assessment, Root Cause, and Effectiveness

24 Safety, Standards, and Professional Conduct

Electrical, Pressure, Chemical, and Laboratory Hazards PPE, Safety Data Sheets, 6S, and Safe Restoration Quality Standards, GUM, and VIM Accreditation Scope and Professional Ethics

Propagation, Covariance, and Effective Degrees of Freedom

Key Takeaways Explain This Topic AI

Uncorrelated standard uncertainty contributions combine by root sum square after sensitivity scaling.

Correlated inputs require covariance terms rather than an unqualified RSS calculation.

Welch–Satterthwaite estimates effective degrees of freedom under its component and model assumptions.

Last updated: October 2026

Once all individual standard uncertainties—both Type A and Type B—have been quantified and converted into equivalent standard deviations ( u ( x i ) u(x_i) u ( x i ​ ) ), the metrologist must combine them into a single comprehensive metric: the combined standard uncertainty ( u c ( y ) u_c(y) u c ​ ( y ) ) .

This synthesis is governed by GUM Clause 5.1, known formally as the Law of Propagation of Uncertainty .

The Law of Propagation of Uncertainty

In metrology, the measurand Y Y Y is rarely measured directly; it is typically determined from N N N other input quantities X 1 , X 2 , … , X N X_1, X_2, \dots, X_N X 1 ​ , X 2 ​ , … , X N ​ through a defined functional relationship: Y = f ( X 1 , X 2 , … , X N ) Y = f(X_1, X_2, \dots, X_N) Y = f ( X 1 ​ , X 2 ​ , … , X N ​ )

Sensitivity Coefficients ( c i c_i c i ​ )

Each input quantity X i X_i X i ​ affects the output Y Y Y according to its partial derivative, termed the sensitivity coefficient ( c i c_i c i ​ ) : c i = ∂ f ∂ x i = ∂ f ∂ X i ∣ X 1 = x 1 , X 2 = x 2 , … , X N = x N c_i = \frac{\partial f}{\partial x_i} = \left. \frac{\partial f}{\partial X_i} \right|_{X_1 = x_1, X_2 = x_2, \dots, X_N = x_N} c i ​ = ∂ x i ​ ∂ f ​ = ∂ X i ​ ∂ f ​ ​ X 1 ​ = x 1 ​ , X 2 ​ = x 2 ​ , … , X N ​ = x N ​ ​

The sensitivity coefficient describes how much the output estimate y y y changes per unit change in input estimate x i x_i x i ​ . The uncertainty contribution from component i i i to the output is: u i ( y ) = ∣ c i ∣ ⋅ u ( x i ) u_i(y) = |c_i| \cdot u(x_i) u i ​ ( y ) = ∣ c i ​ ∣ ⋅ u ( x i ​ )

The Root Sum Square (RSS) Method for Uncorrelated Inputs

When all input quantities are mutually uncorrelated in the stated first-order model, the first-order Taylor series approximation of the combined variance u c 2 ( y ) u_c^2(y) u c 2 ​ ( y ) is the sum of the squared individual contributions: u c 2 ( y ) = ∑ i = 1 N ( ∂ f ∂ x i ) 2 u 2 ( x i ) = ∑ i = 1 N c i 2 u 2 ( x i ) = ∑ i = 1 N u i 2 ( y ) u_c^2(y) = \sum_{i=1}^N \left( \frac{\partial f}{\partial x_i} \right)^2 u^2(x_i) = \sum_{i=1}^N c_i^2 u^2(x_i) = \sum_{i=1}^N u_i^2(y) u c 2 ​ ( y ) = i = 1 ∑ N ​ ( ∂ x i ​ ∂ f ​ ) 2 u 2 ( x i ​ ) = i = 1 ∑ N ​ c i 2 ​ u 2 ( x i ​ ) = i = 1 ∑ N ​ u i 2 ​ ( y )

Taking the square root yields the celebrated Root Sum Square (RSS) formula: u c ( y ) = ∑ i = 1 N [ c i ⋅ u ( x i ) ] 2 = u 1 2 ( y ) + u 2 2 ( y ) + ⋯ + u N 2 ( y ) u_c(y) = \sqrt{\sum_{i=1}^N [c_i \cdot u(x_i)]^2} = \sqrt{u_1^2(y) + u_2^2(y) + \dots + u_N^2(y)} u c ​ ( y ) = i = 1 ∑ N ​ [ c i ​ ⋅ u ( x i ​ ) ] 2 ​ = u 1 2 ​ ( y ) + u 2 2 ​ ( y ) + ⋯ + u N 2 ​ ( y ) ​

Correlated Inputs and Covariance Terms

When two or more input quantities are physically or metrologically dependent, the simple RSS formula is insufficient. Examples of correlation include:

  • Two reference instruments calibrated against the same master standard during the same calibration run
  • Two temperature probes sharing the same digital readout thermometer

Gage blocks evaluated using the same mechanical comparator whose thermal expansion drifts with ambient room temperature

The General Propagation Equation with Covariance

Under GUM Clause 5.2, when correlation exists between input quantities X i X_i X i ​ and X j X_j X j ​ , the combined variance must include cross-product covariance terms: u c 2 ( y ) = ∑ i = 1 N c i 2 u 2 ( x i ) + 2 ∑ i = 1 N − 1 ∑ j = i + 1 N c i c j u ( x i , x j ) u_c^2(y) = \sum_{i=1}^N c_i^2 u^2(x_i) + 2 \sum_{i=1}^{N-1} \sum_{j=i+1}^N c_i c_j u(x_i, x_j) u c 2 ​ ( y ) = i = 1 ∑ N ​ c i 2 ​ u 2 ( x i ​ ) + 2 i = 1 ∑ N − 1 ​ j = i + 1 ∑ N ​ c i ​ c j ​ u ( x i ​ , x j ​ )

Where u ( x i , x j ) u(x_i, x_j) u ( x i ​ , x j ​ ) is the estimated covariance, defined in terms of the correlation coefficient r ( x i , x j ) r(x_i, x_j) r ( x i ​ , x j ​ ) : u ( x i , x j ) = r ( x i , x j ) ⋅ u ( x i ) ⋅ u ( x j ) , where − 1 ≤ r ( x i , x j ) ≤ + 1 u(x_i, x_j) = r(x_i, x_j) \cdot u(x_i) \cdot u(x_j), \quad \text{where } -1 \le r(x_i, x_j) \le +1 u ( x i ​ , x j ​ ) = r ( x i ​ , x j ​ ) ⋅ u ( x i ​ ) ⋅ u ( x j ​ ) , where − 1 ≤ r ( x i ​ , x j ​ ) ≤ + 1

Metrological Impact of Correlation

Consider a differential measurement where Y = X 1 − X 2 Y = X_1 - X_2 Y = X 1 ​ − X 2 ​ ( c 1 = + 1 , c 2 = − 1 c_1 = +1, c_2 = -1 c 1 ​ = + 1 , c 2 ​ = − 1 ): u c 2 ( y ) = u 2 ( x 1 ) + u 2 ( x 2 ) − 2 r ( x 1 , x 2 ) u ( x 1 ) u ( x 2 ) u_c^2(y) = u^2(x_1) + u^2(x_2) - 2 r(x_1, x_2) u(x_1) u(x_2) u c 2 ​ ( y ) = u 2 ( x 1 ​ ) + u 2 ( x 2 ​ ) − 2 r ( x 1 ​ , x 2 ​ ) u ( x 1 ​ ) u ( x 2 ​ )

Positive Correlation ( r > 0 r > 0 r > 0 ): If both instruments drift in the same direction due to a common temperature change, the covariance term is subtracted , reducing the combined uncertainty! In the limit where u ( x 1 ) = u ( x 2 ) u(x_1) = u(x_2) u ( x 1 ​ ) = u ( x 2 ​ ) and r = + 1 r = +1 r = + 1 , the common-mode uncertainty completely cancels ( u c ( y ) = 0 u_c(y) = 0 u c ​ ( y ) = 0 ).

Negative Correlation ( r < 0 r < 0 r < 0 ): If the errors move in opposite directions, the negative signs multiply to form a positive addition, increasing the combined uncertainty beyond the uncorrelated RSS value.

Effective Degrees of Freedom: The Welch-Satterthwaite Equation

Degrees of freedom can be added for appropriate pooled independent variance estimates, but not arbitrarily for every uncertainty model. However, when combining variances from distributions with vastly different magnitudes and degrees of freedom, simple addition fails.

To determine the reliability of the combined standard uncertainty u c ( y ) u_c(y) u c ​ ( y ) , metrologists compute the effective degrees of freedom ( ν eff \nu_{\text{eff}} ν eff ​ ) using the Welch-Satterthwaite equation : ν eff = u c 4 ( y ) ∑ i = 1 N u i 4 ( y ) ν i = u c 4 ( y ) ∑ i = 1 N [ c i u ( x i ) ] 4 ν i \nu_{\text{eff}} = \frac{u_c^4(y)}{\sum_{i=1}^N \frac{u_i^4(y)}{\nu_i}} = \frac{u_c^4(y)}{\sum_{i=1}^N \frac{[c_i u(x_i)]^4}{\nu_i}} ν eff ​ = ∑ i = 1 N ​ ν i ​ u i 4 ​ ( y ) ​ u c 4 ​ ( y ) ​ = ∑ i = 1 N ​ ν i ​ [ c i ​ u ( x i ​ ) ] 4 ​ u c 4 ​ ( y ) ​

Where:

u c ( y ) u_c(y) u c ​ ( y ) is the combined standard uncertainty

u i ( y ) = c i u ( x i ) u_i(y) = c_i u(x_i) u i ​ ( y ) = c i ​ u ( x i ​ ) is the individual uncertainty contribution from input i i i

  • ν i \nu_i ν i ​ is the degrees of freedom associated with component i i i
  • Determining Degrees of Freedom ( ν i \nu_i ν i ​ ) for Budget Components

Type A Components: For a series of n n n repeated measurements, ν i = n − 1 \nu_i = n - 1 ν i ​ = n − 1 . For pooled standard deviations, ν i = ∑ ( n j − 1 ) \nu_i = \sum (n_j - 1) ν i ​ = ∑ ( n j ​ − 1 ) .

Type B Components (Standard Practice): When a Type B limit is derived from a limit or uncertainty estimate whose reliability justifies negligible uncertainty in that estimate (do not assume this from an NMI name alone), the relative uncertainty of the uncertainty is essentially zero, and the degrees of freedom are taken as infinite: ν i → ∞ ⟹ u i 4 ν i = u i 4 ∞ = 0 \nu_i \to \infty \implies \frac{u_i^4}{\nu_i} = \frac{u_i^4}{\infty} = 0 ν i ​ → ∞ ⟹ ν i ​ u i 4 ​ ​ = ∞ u i 4 ​ ​ = 0

Type B Components with Subjective Limits: If a technician estimates a Type B limit with an estimated relative uncertainty R = Δ u ( x i ) / u ( x i ) R = \Delta u(x_i) / u(x_i) R = Δ u ( x i ​ ) / u ( x i ​ ) (for example, 25% relative uncertainty in the estimated standard uncertainty), GUM formula G.3 provides: ν i ≈ 1 2 ( Δ u ( x i ) u ( x i ) ) − 2 = 1 2 ( 0.25 ) 2 = 1 2 × 0.0625 = 8 \nu_i \approx \frac{1}{2} \left( \frac{\Delta u(x_i)}{u(x_i)} \right)^{-2} = \frac{1}{2 (0.25)^2} = \frac{1}{2 \times 0.0625} = 8 ν i ​ ≈ 2 1 ​ ( u ( x i ​ ) Δ u ( x i ​ ) ​ ) − 2 = 2 ( 0.25 ) 2 1 ​ = 2 × 0.0625 1 ​ = 8

Important

Welch-Satterthwaite Truncation Rule: Under GUM Annex G, the computed effective degrees of freedom ν eff \nu_{\text{eff}} ν eff ​ can be truncated down to the nearest integer when using an integer-only t table conservatively (never rounded up) to maintain a conservative estimate of coverage.

Worked covariance comparison

Suppose two input estimates each have the same output units, with standard uncertainties of 3 units and 4 units. If the output is their difference and they are uncorrelated, variance is nine plus sixteen, giving a standard uncertainty of five units. With a supported correlation coefficient of positive one-half, the covariance term subtracts twelve square units. Variance becomes thirteen and the standard uncertainty is about 3.606 units.

If the output were the sum, both sensitivity coefficients would be positive. The same covariance would add twelve square units, producing variance thirty-seven and standard uncertainty about 6.083 units. Correlation therefore cannot be described as always increasing or always reducing uncertainty; its effect depends on both the correlation and the signs of the sensitivities.

Perfect cancellation in an ideal difference applies only to the modeled common contribution. Independent resolution, repeatability, imperfect matching, or other residual effects can remain. Establish covariance from suitable information rather than assuming that instruments from the same supplier share identical errors. Ordinary component-wise Welch–Satterthwaite calculations also need review when correlated variance estimates are involved.