# Read error bars in Book Isle: identify SD, SE and confidence intervals first

Language: en
Canonical: [Read error bars in Book Isle: identify SD, SE and confidence intervals first](<https://allluckly.cn/en/guides/nasbook-error-bar-caption-meaning/>)
中文：[中文 HTML](<https://allluckly.cn/guides/nasbook-error-bar-caption-meaning/>)
English: [English HTML](<https://allluckly.cn/en/guides/nasbook-error-bar-caption-meaning/>)
Published: 2026-11-07T00:00:00+08:00
Modified: 2026-11-07T09:41:37+08:00
Author: Bison

Use captions, sample definitions and an invented 10±2 example to record bar meaning and endpoints without declaring significance from overlap alone.

English translation of the [Chinese original](<https://allluckly.cn/guides/nasbook-error-bar-caption-meaning/>), translated on 2026-11-07. Original publication dates, attribution, code and version notes are retained.

A capped line may represent standard deviation, standard error, a confidence interval or another range; similar appearance does not mean identical meaning. Use Book Isle to retain figure, caption and methods locations and create a manual reading card. All examples are invented and examine quantitative representation, not real study or medical conclusions. Reading, search and notes do not imply automatic error-bar recognition, statistical testing or completion of unstated methods.

## Keep the figure and caption under one identity

Choose a legitimately readable figure with clear axes and caption. Record title, edition, chapter, figure, subfigure letter and a nearby phrase. Retain printed and file-page numbers for PDF or chapter anchors for reflow. Do not crop only bar tops and whiskers: type, units and sample count may appear outside the plot or in methods. Identify whether the centre is a mean, column top or another statistic instead of automatically calling every bar an average. Verify blurred material against the same source or a clearer authorised copy rather than guessing exact endpoints from caps. Begin with identifiable evidence, not reconstruction by imagination.

## Record type from the author, not visual shape

Find the caption’s actual SD, SE, SEM, CI, range or minimum/maximum wording, retaining percentiles, confidence level and sidedness where given. Standard deviation concerns spread, standard error variation of an estimator, and confidence intervals depend on their method; names are not interchangeable. If only error bars is stated, mark type unspecified and consult methods or definitions instead of choosing a familiar interpretation. Subfigures may use different conventions, so attach scope to each. Recheck lost letters in copied abbreviations rather than interpreting SE miscopied as SD. Establish meaning first, then length.

## Record centre and both endpoints separately

On an invented linear vertical axis in exercise units, let a mean be 10 with a caption explicitly giving SD arms of 2 units each. Endpoints are 8 and 12. Retain centre 10, lower 8, upper 12 and SD type, not an unexplained 10±2. Full span is 4; adding the two arms does not replace the centre with 4. Record asymmetric endpoints separately instead of forcing a common plus/minus value. Preserve approximation and reading range where screen evidence cannot support decimals. These invented values make no claim about an actual dataset.

## Identify what the sample count actually counts

Check whether n counts independent individuals, experiments, repeated measurements of one object or another unit, retaining the author’s explanation. Ten readings are not automatically ten independent samples; wells or photographs do not become independent repetitions by count alone. Mark absent n as missing rather than inferring it from visible points that may overlap. Separate description from inferential conditions instead of assuming large counts validate every method. Personal fields for sampling unit, independence statement and unknowns are not a new statistics form in Book Isle. Keep missing information explicitly unresolved, not zero-filled.

## Recalculate the SE relation only under stated conditions

If the invented example additionally states SD=2, n=4 independent observations and the usual standard-error-of-the-mean relation, compute SE=2/√4=1. Write these conditions first and do not silently rename the plotted SD bars SE bars. SE=1 also does not automatically make a 95% CI 10±2; interval method, distribution assumptions and sampling conditions still need checking. This is not a formal test, but shows that one centre may have differently meaningful bars. Retain other estimators or dependence structures used by the source rather than imposing this formula. Extra calculator digits do not supply missing assumptions.

## Separate visual overlap from significance

For two groups, record centres, each bar type and endpoints, then whether they visually overlap. Overlap alone does not establish no difference, and non-overlap alone does not establish significance under every condition. Pairing, dependence, target comparison and method matter; consult the formal analysis. Attach any P value or interval to its stated comparison rather than extending one star across the whole plot. Initially describe interval intersection only, without inferring cause, effect magnitude or practical importance. Keep observation separate from inference and leave unsupported claims unverified; a polished figure cannot replace methods.

## Return to the figure and check exported symbols

Bookmark or externally locate figure, caption and methods, leave the reader and return to each, confirming edition and subfigure. Hide your endpoint result and reread so memory does not drive the reading. Reopen copied/exported notes and check plus/minus, square root, endpoints and SD/SE letters, retaining unspecified labels. Plain prose may replace difficult symbols without dropping per-arm versus full-span meaning. Rematch figure and caption after reflow or revision rather than relying on old pages. Explain corrections, distinguishing symbol transcription from a changed understanding of type.

## Finish the card with a bounded conclusion

Retain source, centre statistic, bar type, endpoints, sample definition and unverified inferences. Explain why equal lengths can mean different things, unknown n prevents casual conversion and overlap alone does not decide significance. If caption meaning remains absent, state that statistical interpretation is unresolved rather than adding defaults. Identify the next figure’s convention afresh. Keep the original record and one specific question. The card establishes the extent of your checking, not correctness of an entire study or automatic chart analysis by Book Isle.

## Related tutorials and sources

[Read logarithmic plots in Book Isle: interpret equal distances as ratios, not differences](<https://allluckly.cn/en/guides/nasbook-log-axis-ratio-reading/>); [Read experiment descriptions in Book Isle: separate factors, responses and uncontrolled conditions](<https://allluckly.cn/en/guides/nasbook-experiment-factor-response-reading/>).

Sources checked on 7 November 2026: the [official Book Isle App Store listing](<https://apps.apple.com/cn/app/id6766198898>) and [Apple Lookup](<https://itunes.apple.com/lookup?id=6766198898&country=cn>) show version 1.1.4, requiring iOS / iPadOS 15.6 or later. This workflow uses local reading, search, bookmarks and notes for manual checking. Readers record the samples and conclusions; automatic chart recognition, statistical testing or proof verification is not assumed. Confirm available controls against the installed version and file type. The exercise does not claim step-by-step device testing in this batch.

Method reference: [Cumming, Fidler and Vaux: Error bars in experimental biology](<https://pmc.ncbi.nlm.nih.gov/articles/PMC2064100/>) discusses types and interpretation limits. This exercise does not apply its particular test-approximation rules; values and reading records are invented.