The Engineer’s 9-Step Uncertainty Analysis Checklist

This checklist provides a systematic framework for conducting a thorough uncertainty analysis. It begins with the strategic purpose (Step 0) and moves through the tactical execution of the calculation, providing technical details, engineering insights, and exam tips at each stage.

Step 0: Define Purpose, Context, and Requirements¶

Objective: To establish why the measurement is being performed and what “good enough” means. This step defines the required precision and the governing rules before any work begins.

Technical Checklist:

• Define the Purpose: Is this for...

  • Basic Science: Verifying a hypothesis?

  • Quality Control: Checking if a manufactured part meets its design specifications?

  • Regulatory Compliance: Proving that a product meets a legal standard (e.g., emissions)?

• Determine Required Uncertainty: What is the target uncertainty? This is often defined by a tolerance. The uncertainty of your measurement should ideally be significantly smaller than the tolerance (e.g., a 4:1 or 10:1 ratio is often desired).

• Identify Governing Standards: Are there any external standards that must be followed?

  • ISO/IEC 17025: For accredited calibration and testing laboratories.

  • GUM (Guide to the Expression of Uncertainty in Measurement): The foundational document.

  • Industry-specific standards (e.g., ASTM, ASME).

Explanation & Engineering Insight: This step prevents the two most common measurement mistakes: 1) Over-engineering: spending excessive time and money to get a highly precise result when it’s not needed, and 2) Under-engineering: getting a result that is useless because its uncertainty is too large to make a meaningful decision. This is the “fitness for purpose” step.

How this appears on an exam (MCQ Tips):

• Keywords: “tolerance,” “specification,” “fitness for purpose,” “compliance.”

• Typical Question: “A specification requires a shaft to be $10.00 \pm 0.04$ mm. To confidently check for compliance, the measurement’s expanded uncertainty ($U$) should be...”

  • Correct Answer: Significantly smaller than 0.04 mm.

  • Distractors: Equal to 0.04 mm, Larger than 0.04 mm.

• Tip: The measurement is only useful if its uncertainty is small enough to resolve the tolerance.

Step 1: Develop the Measurement Model¶

Objective: To create the mathematical link between the final quantity you want (the measurand) and the inputs you will actually measure.

Technical Checklist:

• Identify the Measurand ($Y$): What is the final quantity to be calculated? (e.g., Density, $\rho$).

• List All Input Quantities ($X_1, X_2, \dots, X_N$): What variables will you measure directly? (e.g., mass $m$, length $L$, diameter $D$).

• Write the Model Equation: Formulate the equation $Y = f(X_1, X_2, \dots, X_N)$.

  • Example: For density of a cylinder, $\rho = \frac{m}{V} = \frac{m}{\pi (D/2)^2 L}$.

Explanation & Engineering Insight: The model is the blueprint for your entire calculation. Every variable in this equation will become a source of uncertainty that you must analyze. If a variable is missing from the model (e.g., you forget to include a temperature correction factor), its uncertainty is ignored, leading to an incorrect final result.

How this appears on an exam (MCQ Tips):

• Keywords: “measurand,” “input quantity,” “measurement model.”

• Typical Question: “To determine the kinetic energy ($E = \frac{1}{2}mv^2$) of an object, what are the input quantities?”

  • Correct Answer: Mass ($m$) and velocity ($v$).

  • Distractors: Energy (it’s the measurand), constants like $\frac{1}{2}$.

• Tip: The input quantities are the variables you physically measure.

Step 2: Identify and Characterize All Uncertainty Sources¶

Objective: To brainstorm and list every possible source of error for each input quantity ($X_i$).

Technical Checklist:

• Identify Type A Sources (Random): This is typically the repeatability of your measurement, found by taking multiple readings.

• Identify Type B Sources (Systematic): Systematically list all “hidden” uncertainties for each input.

  • Instrument: Calibration uncertainty, resolution, hysteresis.

  • Environment: Temperature, humidity, vibration.

  • Operator: Parallax, technique variations.

  • Assumptions: Uncertainty in any constants or assumptions in your model.

Explanation & Engineering Insight: This is the detective work. A common failure is to focus only on repeatability (Type A) while ignoring the instrument’s inherent limitations (Type B). A precise-looking result can be deeply flawed if based on an uncalibrated tool.

How this appears on an exam (MCQ Tips):

• Keywords: “Type A evaluation,” “Type B evaluation,” “systematic,” “random.”

• Typical Question: “The uncertainty arising from the limited resolution of a digital display is determined by...”

  • Correct Answer: A Type B evaluation.

  • Distractors: A Type A evaluation, Repeatability, Statistical analysis of multiple readings.

• Tip: If you didn’t find the uncertainty by calculating the standard deviation of your own repeated measurements, it’s a Type B evaluation.

Step 3: Quantify Standard Uncertainties ($u_i$)¶

Objective: To assign a numerical value to each uncertainty source and convert it into a standard uncertainty ($u$), which is equivalent to one standard deviation.

Technical Checklist:

• Quantify Type A: For $n$ repeated measurements with standard deviation $s$, the standard uncertainty of the mean is:

  $$u = \frac{s}{\sqrt{n}}$$

  (1)

• Quantify Type B: Use the appropriate divisor based on the assumed probability distribution. For a range of $\pm a$:

  • Rectangular Distribution: (You only know the bounds are absolute).

    $$u = \frac{a}{\sqrt{3}}$$

    (2)

  • Normal Distribution: (You have a 95% confidence value).

    $$u = \frac{a}{2} \quad (\text{or more precisely, } \frac{a}{1.96})$$

    (3)

  • Triangular Distribution: (You know values are more likely near the center).

    $$u = \frac{a}{\sqrt{6}}$$

    (4)

Explanation & Engineering Insight: This step standardizes all uncertainties into a common “currency.” The divisor is critical—it converts a maximum bound or a confidence interval into a 1-sigma equivalent, which is necessary for the next step.

How this appears on an exam (MCQ Tips):

• Typical Question: “A manufacturer’s specification sheet states a power supply is accurate to $5.0 \pm 0.1$ V. Assuming a rectangular probability distribution, what is the standard uncertainty?”

  • Correct Answer: $0.1 / \sqrt{3}$ V.

  • Distractors: 0.1 V, $0.1 / 2$ V, $0.1 / \sqrt{6}$ V.

• Tip: Memorize the rectangular ($\sqrt{3}$) and normal (2) divisors. They are the most common.

Step 4: Check for Correlations¶

Objective: To determine if the uncertainty sources for different input variables are independent.

Technical Checklist:

• Identify Common Influences: Was the same instrument (e.g., thermometer), standard, or procedure used to measure multiple input quantities ($X_i$ and $X_j$)?

• State Assumption: If no correlations are found, explicitly state that all inputs are assumed to be uncorrelated.

• Calculate Covariance (Advanced): If sources are correlated, the covariance $u(x_i, x_j)$ must be calculated and used in the full uncertainty formula.

Explanation & Engineering Insight: For most introductory work, you can assume no correlation, but it’s a professional requirement to show you’ve considered it. Ignoring a strong correlation can lead to a significant underestimate of the total uncertainty.

How this appears on an exam (MCQ Tips):

• Typical Question: “Correlation between the uncertainties of two input quantities is most likely to occur when...”

  • Correct Answer: They are measured using the same calibrated instrument.

  • Distractors: They are measured in different labs, at different times, with different instruments.

• Tip: Look for a single shared source of potential error affecting multiple inputs.

Step 5: Calculate Combined Standard Uncertainty ($u_c$)¶

Objective: To combine all individual standard uncertainties into a single total standard uncertainty for the final measurand.

Technical Checklist:

• Calculate Sensitivity Coefficients ($c_i$): Determine the partial derivative of the model with respect to each input: $c_i = \frac{\partial f}{\partial X_i}$. This quantifies how much the output ($y$) changes for a small change in an input ($x_i$).

• Apply the Law of Propagation of Uncertainty (RSS): For uncorrelated inputs, the formula is:

  $$u_c(y) = \sqrt{\sum_{i=1}^{N} [c_i u(x_i)]^2} = \sqrt{[c_1 u(x_1)]^2 + [c_2 u(x_2)]^2 + \dots + [c_N u(x_N)]^2}$$

  (5)

Explanation & Engineering Insight: This is “error propagation.” The Root-Sum-of-Squares (RSS) method provides a realistic combination of uncertainties, based on the statistical likelihood that not all errors will be at their maximum value simultaneously.

How this appears on an exam (MCQ Tips):

• Typical Question: “A calculation has two uncorrelated standard uncertainties, $u_1 = 6$ and $u_2 = 8$. What is the combined standard uncertainty, $u_c$?”

  • Correct Answer: $\sqrt{6^2 + 8^2} = \sqrt{36 + 64} = \sqrt{100} = 10$.

  • Distractors: 14 (incorrectly adding), 7 (incorrectly averaging).

• Tip: If asked to combine uncertainties, immediately think “Pythagorean theorem” or RSS. The simple sum is the most common incorrect choice.

Step 6: Calculate Expanded Uncertainty ($U$)¶

Objective: To scale the result to an interval with a higher level of confidence (typically 95%).

Technical Checklist:

• Choose a Coverage Factor ($k$): For most situations, a 95% confidence level is desired, so use $k=2$.

• Calculate Expanded Uncertainty:

  $$U = k \times u_c(y)$$

  (6)

Explanation & Engineering Insight: The combined standard uncertainty ($u_c$) only gives ~68% confidence. This is insufficient for most engineering work. Multiplying by $k=2$ expands this interval to ~95% confidence, which is the industry standard for reporting.

How this appears on an exam (MCQ Tips):

• Keywords: “Expanded Uncertainty,” “Coverage Factor,” “95% confidence.”

• Typical Question: “A measurement’s combined standard uncertainty is $u_c = 0.3$ V. The expanded uncertainty $U$ for a 95% confidence level is...”

  • Correct Answer: $2 \times 0.3 \text{ V} = 0.6 \text{ V}$.

  • Distractors: 0.3 V, 0.15 V.

• Tip: Remember the progression: Standard ($u_i$) $\rightarrow$ Combined ($u_c$) $\rightarrow$ Expanded ($U$).

Step 7: Create the Uncertainty Budget¶

Objective: To present all your analysis in a clear, organized table.

Technical Checklist:

• Construct a Table: Create columns for:

  1. Source of Uncertainty (e.g., “Mass measurement repeatability”)

  2. Value (±a)

  3. Distribution (e.g., “Normal”)

  4. Divisor (e.g., “2”)

  5. Standard Uncertainty ($u_i$)

  6. Sensitivity Coefficient ($c_i$)

  7. Contribution to Variance $(c_i u_i)^2$

• Sum the Final Column: The square root of the total of the final column is your $u_c(y)$. This is a good self-check.

Explanation & Engineering Insight: The budget is your proof of work. It transparently shows every step of your analysis. Crucially, it identifies the dominant uncertainty source—the largest value in the final column. If you need to improve your measurement, this tells you exactly where to focus your effort.

How this appears on an exam (MCQ Tips):

• Typical Question: “According to the uncertainty budget provided, which input contributes most to the final uncertainty?”

  • Correct Answer: Look for the largest value in the Contribution to Variance column.

  • Distractors: Other sources that might have a large initial uncertainty but a small sensitivity coefficient.

• Tip: The biggest contributor is not always the source with the largest standard uncertainty ($u_i$), but the one with the largest weighted contribution ($(c_i u_i)^2$).

Step 8: Report the Final Result¶

Objective: To communicate your final answer in a standardized, unambiguous format.

Technical Checklist:

• Use the Standard Format:

  $$\text{Result} = y \pm U \; (\text{units})$$

  (7)

• State the Confidence Level: Always include a statement like:

  “The reported uncertainty is an expanded uncertainty, calculated using a coverage factor of $k=2$, which gives a level of confidence of approximately 95%.”

Explanation & Engineering Insight: This final statement is the universal language of measurement. It provides the best estimate ($y$), the range of doubt ($U$), and the confidence in that range ($k=2$, 95%). Any engineer can look at this result and know exactly what it means and how to use it in their own work.

How this appears on an exam (MCQ Tips):

• Typical Question: “A measurement gives a final value of $y = 25.4$ mm and a combined standard uncertainty of $u_c = 0.2$ mm. What is the correctly reported result at a 95% confidence level?”

  • Correct Answer: $25.4 \pm 0.4$ mm.

  • Distractors: $25.4 \pm 0.2$ mm (used $u_c$ not $U$), $25.4 \pm 0.1$ mm (incorrect calculation).

• Tip: Always use the Expanded Uncertainty ($U$) for the final ± value.