Chapter 4 of 8
Chapter 4: Measurement Bias and Confidence Intervals
4.1 Introduction
Up to this point, measurement in MET 135 has focused on how to measure using different instruments. In industry, however, knowing how to measure is not enough. Engineers and technicians must also determine whether a measurement system can be trusted.
This chapter introduces measurement bias and confidence intervals, two core statistical tools used to evaluate the performance of measurement systems. These concepts are widely used in manufacturing, quality control, and process improvement to decide whether an instrument–operator system is acceptable.
This chapter connects hands-on metrology with data analysis using Excel and formal decision-making rules.
4.2 What Is Measurement Bias?
Bias is defined as a statistically significant difference between measured values and the true or accepted reference value of a gage.
Bias can result from:
- Instrument wear or miscalibration
- Improper measurement technique
- Inconsistent measuring force
- Environmental conditions
- Operator habits
Bias does not mean a single bad measurement. Bias refers to a systematic shift in the measurement system.
4.3 Reference Values and Targets
To evaluate bias, measurements must be compared to a known reference:
- Gage block values
- Expert measurements
- Certified reference dimensions
In bias studies, the target value is zero, meaning no difference between measured and actual values.
Bias = Measured Value − Actual Value
4.4 Why We Use Confidence Intervals
When measuring large populations of parts, it is rarely practical to measure every part. Instead, engineers measure small samples and use statistics to estimate the population behavior.
A confidence interval (CI) provides a range of values that the true population mean is likely to fall within, based on:
- Sample mean
- Sample standard deviation
- Sample size
- Selected confidence level
Confidence intervals acknowledge uncertainty while still enabling decisions.
4.5 Confidence Level and Risk
Common confidence levels include:
- 90% (α = 0.10)
- 95% (α = 0.05)
- 99% (α = 0.01)
A 95% confidence level means that 95 out of 100 similar studies would produce intervals containing the true mean.
Higher confidence means less risk — but wider intervals.
4.6 Student's t-Statistic
Because measurement studies typically use small sample sizes, Student's t-distribution is used instead of the normal distribution.
Key parameters:
- Degrees of freedom (df) = n − 1
- Selected confidence level
- t-statistic value from table or Excel
Excel function:
T.INV(probability, degrees_freedom)
4.7 Confidence Interval Formula
The confidence interval for the mean is calculated as:
x̄ ± (t · s / √n)
Where:
- x̄ = sample mean
- t = Student's t-statistic
- s = sample standard deviation
- n = number of trials
The upper and lower confidence limits are:
UCL = x̄ + (t · s / √n)
LCL = x̄ − (t · s / √n)
4.8 Interpreting Bias Using Confidence Intervals
For bias studies:
- If the confidence interval straddles zero (one limit positive, one negative): → No evidence of bias
- If both limits are positive or both negative: → Bias is present
Important rule: Do not use the sample mean (x̄) alone to decide bias. Decisions are based only on UCL and LCL.
Try It: Confidence Interval & Bias Calculator
Enter your sample statistics. The tool looks up the two-tailed t-statistic, computes both limits, and applies the bias decision rule.
4.9 Excel Integration
Excel is used to:
- Calculate trial means
- Compute sample standard deviation
- Apply the confidence interval formula
- Automate t-statistic selection
Using Excel ensures consistency and mirrors industry practice.
4.10 Chapter Summary — Study Points
- Bias is a systematic measurement error
- Bias compares measurements to a known reference
- Confidence intervals estimate population behavior
- Student's t-statistic is used for small samples
- Bias decisions depend on UCL and LCL only
- Excel is a required analysis tool
Chapter 4 Exercises
Each exercise includes a worked example followed by a student practice problem.
Exercise 1: Calculating Bias
Example (Solved)
Actual diameter = 0.5000 in
Measured diameter = 0.5003 in
Bias = 0.5003 − 0.5000 = +0.0003 in
Student Practice
Actual diameter = 12.50 mm
Measured diameter = 12.47 mm
Calculate the bias.
Show answer
Bias = 12.47 − 12.50 = −0.03 mm
Exercise 2: Degrees of Freedom
Example (Solved)
Number of trials = 8
Degrees of freedom: df = 8 − 1 = 7
Student Practice
A study uses 12 trials. Determine the degrees of freedom.
Show answer
df = 12 − 1 = 11
Exercise 3: Confidence Interval Calculation
Example (Solved)
x̄ = −0.0001 in
s = 0.00011 in
n = 7
CL = 99%
t (df = 6, 99%) = 3.707
CI = −0.0001 ± (3.707 × 0.00011 / √7) = −0.0001 ± 0.00015
LCL ≈ −0.0003 in
UCL ≈ +0.0001 in
The interval straddles zero, so there is no evidence of bias. (Corrected in this edition: the UCL was previously misprinted as 0.0000 in.)
Student Practice
x̄ = 0.0002 mm
s = 0.00015 mm
n = 10
CL = 95%
Calculate the confidence interval.
Show answer
t (df = 9, 95%) = 2.262
CI = 0.0002 ± (2.262 × 0.00015 / √10) = 0.0002 ± 0.0001
LCL ≈ +0.0001 mm, UCL ≈ +0.0003 mm
Both limits are positive, so bias is present.
Exercise 4: Bias Decision
Example (Solved)
LCL = −0.002 mm
UCL = +0.001 mm
Conclusion: No bias (interval straddles zero)
Student Practice
LCL = +0.001 mm
UCL = +0.004 mm
Determine whether bias is present.
Show answer
Bias is present — both limits are positive, so the interval does not straddle zero.
Exercise 5: Instrument System Evaluation
Example (Solved)
A micrometer system has:
LCL = −0.0004 in
UCL = +0.0002 in
Conclusion: Measurement system is not biased
Student Practice
A caliper system has:
LCL = −0.0006 in
UCL = −0.0001 in
Evaluate the system.
Show answer
Bias is present — both limits are negative. The caliper system reads systematically low and should be investigated (calibration, technique, or wear).
Chapter 4 Resources
- Student's t-statistic table and Excel usage (provided in course materials; a 90/95/99% table is built into the calculator above)