LibreOffice Calc · Analysis and modeling · Intermediate
Summarize and sample data with Statistics tools
A worked Calc example to summarize and sample data with Statistics tools, with source checks, changed inputs and recovery.
Documentation-qualified for Calc 26.2; native workflow and persistence checks pending · Editorial standards
Summarize the numeric measure first, then sample complete records so labels and values stay together. A sample is not a smaller population with automatically identical statistics, and a random sample cannot be checked by expecting one fixed set of rows.
Create the practice population
Use LibreOffice Calc 26.2 desktop, English standard menus on Windows/Linux. These procedures come from versioned Help; no native run or persistence test is claimed. Keep an ODS practice copy with a clear output area.
On Population, enter A1:B9:
| ID | Minutes |
|---|---|
| 101 | 2 |
| 102 | 4 |
| 103 | 4 |
| 104 | 6 |
| 105 | 8 |
| 106 | 10 |
| 107 | 12 |
| 108 | 14 |
There are eight records, total 60, mean 7.5, median 7 and mode 4. Minimum 2, maximum 14 and range 12 provide quick checks that the endpoints were included. IDs are numeric for a transparent sampling fixture, but they are identifiers, not a meaningful measurement to summarize.
Generate descriptive statistics
- Choose Data > Statistics > Descriptive Statistics.
- Set Input Range to
$Population.$B$2:$B$9, excluding the ID column and the header. - Set Grouped By to Columns.
- Set Results to to
$Population.$D$1, ensuring the surrounding output area is empty, then confirm. - Compare Count 8, Sum 60, Mean 7.5, Median 7, Mode 4 and Range 12 with the independent figures above.
The sample variance is 18 and sample standard deviation is about 4.242641; the standard error is 1.5. These differ from the population variance 15.75 because the sample statistic uses n −1. Label the interpretation instead of choosing whichever variance looks smaller.
A Count below 8 is a warning to inspect blank cells or numeric-looking text, not evidence that the tool randomly omitted records. Correct data types before accepting the report; see numbers stored as text.
Take a deterministic periodic sample
Choose Data > Statistics > Sampling. Set Input Range to $Population.$A$2:$B$9, now including both ID and Minutes but excluding headings. Set Results to $Population.$H$2. Choose Periodic and Period 2, then confirm.
The source Help’s example selects the second, fourth and subsequent alternate rows. Here the expected records are (102, 4), (104, 6), (106, 10), (108, 14). Verify complete row pairing, not just the minutes column. The sampled mean is 8.5, which differs from the population’s 7.5 because the ordered population makes this periodic sample unrepresentative of some values.
Keep order is automatically selected for periodic sampling. Periodic selection is not randomization and can align badly with a repeating process cycle.
Draw a random sample without replacement
Run Sampling again into a different blank area, such as K2. Choose Random, Sample size 3, and clear With replacement. Set Keep order if you want selected records returned in original order.
Expect exactly three population records with distinct IDs and their original matching Minutes. Do not require a specific trio or mean. A repeated draw may legitimately produce a different trio or the same one. With replacement permits repeated records and can allow samples larger than the population; it answers a different sampling design and must be disclosed when used.
Change, rerun and preserve provenance
Change the final Minutes value from 14 to 18. The full-population sum becomes 64, mean 8 and maximum 18. Regenerate descriptive statistics in a new blank area so old and new results can be compared without ambiguity. Rerun Period 2 sampling: the final sampled record should be (108, 18), and its sample mean becomes 9.5.
Do not assume a previously generated random extract is a continuously maintained view. Record its source range, extraction time and replacement/order settings. Preserve the old sample if it is part of an audit, but label it as old.
Restore 14, regenerate the baseline report, save and reopen the ODS, and confirm original source values plus clearly labeled outputs. If an output overwrote data, undo immediately in the disposable copy. Clear or move the entire old output only after checking its boundaries; avoid leaving stale rows below a smaller new sample.
Official references
Sources were captured on 20 October 2026 and reviewed from retained snapshots. All outcomes above are expected checks, not a claim of native execution.