Imagine drawing many samples of 16 invoices from the same fixed Juniper frame, returning the selected IDs each time, and recomputing the mean. Some samples would contain more short-settlement invoices; others would contain more long ones. Their means would form a sampling distribution around a center.
That variation is not a calculation error. It is the consequence of observing a subset.
Keep three kinds of movement separate
- Sampling variability: different units were selected from the same population under the same design.
- Population or process change: the underlying settlement behavior actually differs across periods or groups.
- Data or design change: the frame, cutoff, definition, selection method, coding, or missingness changed.
Juniper's Year 1 and Year 2 samples come from different year-specific populations. Their 2.5-day difference in sample means could reflect population change and sampling variability, while the 78-day record also raises a tail question. The packet does not identify the cause.
The standard error summarizes expected sample-to-sample variation for a named statistic under a model and design. It does not measure source-system error or ordinary business volatility directly.
Put the concept to work
Understand this concept
- Explain why repeated samples from one fixed population produce a distribution of statistics even when the design and sample size remain unchanged.
Analyze this concept
- Separate plausible sample-to-sample variation from changes in the population, selection design, source system, variable definition, or data quality.
Learning resources
Choose a lesson, try an application, or inspect the sources behind this concept.
Build on these ideas
- Sample — Understand
To understand this concept: Required. Different selected subsets create sampling variability.
- Sampling variability — Understand
To analyze this concept: Required. Diagnosis begins with the mechanism that makes statistics vary under an unchanged design.
- Statistic — Understand
To understand this concept: Required. The quantity that varies across samples is the statistic.
Lessons
Worked examples and cases
Practice
Common mistaken ideas
- Mistaken idea: A 95% confidence level is a 95% probability for this fixed interval
- Mistaken idea: A sample statistic is the population parameter
- Mistaken idea: Random sampling eliminates data error
Show 1 more mistaken ideas
Sources
Related concepts
Show 1 more related concepts
Use this idea next
- Confidence interval — Understand
Required level here: understand. Required. The interval changes because the selected sample and statistic change.
- Sampling variability — Analyze
Required level here: understand. Required. Diagnosis begins with the mechanism that makes statistics vary under an unchanged design.
- Standard error — Understand
Required level here: understand. Required. Standard error quantifies the sampling distribution's spread.