Repeated measurements are observations of the same measurand obtained under defined measurement conditions. They are used to understand how a measurement process behaves and how much variation occurs between observations. A basic analysis calculates the arithmetic mean, median, sample variance and sample standard deviation. The standard uncertainty of the mean can then be estimated as the sample standard deviation divided by the square root of the number of independent observations. However, repeated observations should not be averaged blindly. The measurement process should be sufficiently stable, the measurand should remain unchanged and the observations should be suitable for the statistical model. A time sequence can reveal drift, trends, periodic variation or sudden changes that may be hidden by a single mean and standard deviation. Graphical methods such as time-series plots, histograms, box plots and residual plots help identify patterns. Statistical tests may be used to investigate distributional assumptions, differences between groups or possible outliers, but their assumptions and limitations must be considered. When observations are correlated, the effective amount of independent information may be smaller than the number of recorded measurements. For example, repeated temperature readings taken seconds apart may be strongly related. The analysis should therefore consider independence, stationarity, sampling strategy, measurement conditions and the purpose of the reported result.
Statistical Analysis of Repeated Measurements
Statistical analysis of repeated measurements examines a series of observations to estimate central values, characterize variability, evaluate stability and determine the uncertainty associated with measurement results. It provides the foundation for repeatability studies, Type A uncertainty evaluation and detection of unusual measurement behavior.
For n independent observations, x̄ = Σxᵢ/n, s = √[Σ(xᵢ − x̄)²/(n − 1)] and u(x̄) = s/√n.
Explanation
Variables
| Symbol | Meaning | Si Unit |
|---|---|---|
xᵢ | The i-th repeated measurement | Unit of the measurand |
x̄ | Arithmetic mean of the observations | Same unit as xᵢ |
s | Sample standard deviation | Same unit as xᵢ |
n | Number of observations | Dimensionless count |
u(x̄) | Standard uncertainty of the mean | Same unit as xᵢ |
Example or application
A worked example has not yet been added.
Definition
Statistical analysis of repeated measurements is the application of statistical methods to a series of observations of the same measurand to estimate central tendency, dispersion, stability and measurement uncertainty.
Purpose
To extract reliable information from repeated observations, quantify measurement variability, evaluate process stability and estimate the uncertainty of a reported result.
Importance
Repeated-measurement analysis is central to experimental research, calibration, laboratory testing, manufacturing quality control and scientific data interpretation.
Preparation Of Measurement Data
Confirm that all observations refer to the same measurand.
Record units, instrument identification, measurement times and conditions.
Preserve the original observations.
Check for recording, transcription and unit-conversion errors.
Identify changes in operators, instruments or environmental conditions.
Determine whether the observations are independent or correlated.
Descriptive Statistical Analysis
Central Tendency
Arithmetic mean
Median
Dispersion
Sample variance
Sample standard deviation
Range
Interquartile range
Uncertainty
Standard uncertainty of the mean
Combined standard uncertainty where other contributions are present
Interpretation
Central tendency describes the location of the data, dispersion describes their spread and uncertainty describes the doubt associated with the reported measurement result.
Graphical Analysis Methods
| Method | Purpose |
|---|---|
Time-series plot | Displays observations in measurement order and helps identify drift, trends, cycles and sudden changes. |
Histogram | Shows the distribution and frequency of observed values. |
Box plot | Summarizes the median, quartiles, spread and possible unusual observations. |
Residual plot | Shows differences between observations and a fitted or expected model. |
Scatter plot | Helps identify relationships between measurement results and environmental or experimental variables. |
Statistical Assessment
| Assessment | Description |
|---|---|
Stability | Determine whether the measurement process remains statistically consistent over the observation period. |
Independence | Examine whether successive observations are statistically related. |
Distribution shape | Assess whether the assumed probability distribution is reasonable for the intended analysis. |
Outlier investigation | Investigate unusual observations and determine whether a documented physical, procedural or recording cause exists. |
Trend detection | Determine whether the data contain systematic changes over time. |
Worked Example
Problem
A laboratory records five measurements of a stable reference length: 20.1 mm, 20.2 mm, 20.0 mm, 20.1 mm and 20.1 mm. Calculate the mean, sample standard deviation and standard uncertainty of the mean.
Observations
20.1
20.2
20.0
20.1
20.1
Mean
x̄ = 20.10 mm
Sample Variance
s² = 0.005 mm²
Sample Standard Deviation
s = √0.005 ≈ 0.0707 mm
Standard Uncertainty
u(x̄) = 0.0707/√5 ≈ 0.0316 mm
Result
The mean is 20.10 mm and its Type A standard uncertainty is approximately 0.032 mm.
Interpretation
The uncertainty of the mean is smaller than the standard deviation of individual observations because the mean is estimated from five independent measurements.
Effect Of Sample Size
Description
For independent observations from a stable process with finite variance, the standard uncertainty of the mean decreases approximately in proportion to 1/√n.
Examples
| Sample Size | Relative Standard Error |
|---|---|
n = 1 | 1s |
n = 4 | 0.5s |
n = 9 | 0.333s |
n = 16 | 0.25s |
Limitation
Increasing the number of observations does not remove systematic uncertainty and may provide little benefit when observations are strongly correlated.
Correlated Observations
Description
Observations are correlated when the value of one measurement is statistically related to another.
Examples
Repeated temperature measurements taken close together in time
Measurements affected by slowly changing instrument drift
Repeated measurements using a shared fluctuating reference
Consequence
The simple s/√n formula may underestimate uncertainty if correlation is ignored.
Approach
Use an appropriate time-series, correlation or variance-component model when dependence is significant.
Outlier Handling
Do not remove an observation simply because it differs from the mean.
Check the original record and measurement conditions.
Investigate instrument malfunction, operator mistakes and sample changes.
Use an appropriate statistical test only when its assumptions are satisfied.
Document any exclusion and retain the original data.
Consider robust statistical methods when the data contain genuine extreme values.
Advanced Analysis Methods
| Method | Purpose |
|---|---|
Analysis of variance (ANOVA) | Separates variation associated with different groups or experimental factors. |
Regression analysis | Evaluates relationships between measurements and explanatory variables such as temperature or time. |
Autocorrelation analysis | Examines statistical dependence between observations at different time lags. |
Control charts | Monitor process behavior and identify unusual changes over time. |
Bootstrap analysis | Uses repeated resampling to estimate statistical variability when suitable assumptions and data conditions permit. |
Limitations
A large sample does not guarantee representative measurements.
Averages can hide drift and other time-dependent effects.
Statistical methods cannot correct an incorrectly defined measurand.
Outlier tests may produce misleading results when their assumptions are not met.
Repeated observations do not automatically account for calibration uncertainty or other Type B contributions.
Reference
JCGM 100:2008, Evaluation of Measurement Data — Guide to the Expression of Uncertainty in Measurement (GUM).
ISO 5725 series, Accuracy (Trueness and Precision) of Measurement Methods and Results.
NIST Technical Note 1297, Guidelines for Evaluating and Expressing the Uncertainty of NIST Measurement Results.
Related Topic
SCI-REF-003-083
SCI-REF-003-084
SCI-REF-003-085
SCI-REF-003-090
Sources and verification
Sources are shown as provided in the project data. Not every source has yet been independently verified. Please report errors or missing citations.
- Source verification pending; check an authoritative textbook, university, standards body, or peer-reviewed source.
Suggest a correction or source · More in Units & Measurement