CEPI5215 Chap.5 Tables, Figures and Statistical Reporting
Tables, Figures and Statistical Reporting
Tables and figures are evidence displays with different reader tasks. Tables support retrieval of exact values; figures reveal patterns, distributions and comparisons. This chapter starts with purpose, then checks counts, denominators, units, labels, scales and uncertainty before styling. Statistical reporting emphasises estimates and precision rather than threshold language.
The goal is a standalone display that enables honest comparison without duplicating the same evidence in prose. A display also encodes an argument about what should be compared. Ordering, reference groups, time points and treatment of missing values can direct attention as strongly as colour. Editorial review should therefore test the analytical comparison before refining the graphic form.
What this chapter covers
- 01
Choosing between prose, table and figure from the communication purpose
- 02
Table titles, headings, units, denominators, footnotes and abbreviations
- 03
Figure type, axes, scales, legends and accessible group distinctions
- 04
Effect estimates, confidence intervals and exact p-value reporting
- 05
Avoiding duplication across text, tables and figures
- 06
Consistency checks across abstract, results and display values
Worked example · free
Repair a misleading group comparison
- +1Add the group counts and denominators so the proportions can be interpreted.
- +1Report the prespecified effect measure and confidence interval before any p value.
- +1Use an honest vertical scale or a more informative comparison display that shows absolute magnitude.
- +1Discuss clinical relevance and study limits separately from statistical threshold language.
Key terms
- Denominator
- The total population or observations to which a reported proportion refers.
- Effect estimate
- A quantitative expression of the size and direction of a relationship or difference.
- Confidence interval
- An interval describing estimate precision under the statistical method’s assumptions.
- P value
- A probability calculated under a specified null model, not a measure of effect size or truth.
- Axis truncation
- Starting or ending a graph scale in a way that may exaggerate or suppress visible differences.
- Figure legend
- Explanatory text that identifies symbols, groups and essential information needed to interpret a figure.
Tables, Figures and Statistical Reporting FAQ
When is a table better than a figure?
Choose a table when readers need exact values across multiple variables or categories. Choose a figure when trend, shape, distribution or visual comparison is central. If the message is a single simple value, prose may be more efficient.
Is a non-significant result evidence of no difference?
Not automatically. Examine the estimate and confidence interval, study power, design and assumptions. The data may be compatible with meaningful benefit and harm, or an equivalence framework may be needed to support a claim of similarity.
Can text repeat the table for emphasis?
It should not repeat every value. Use the narrative to state the principal pattern, comparison and interpretation, and direct readers to the display for detail. Repetition consumes attention and can introduce inconsistent numbers during revision.
How many decimal places should be reported?
Use precision appropriate to measurement quality, sample size, uncertainty and journal style. Apply it consistently to comparable values. Extra decimals can imply accuracy the study does not possess, while aggressive rounding can hide meaningful differences.
Why are three-dimensional charts discouraged?
Perspective and surface area distort perceived magnitude while adding no data dimension in ordinary bar or pie displays. A simpler two-dimensional form usually supports more accurate comparison and leaves room for labels and uncertainty.
How should missing data appear in a table?
Identify the analysed denominator and show or explain missing values according to the table purpose. Do not let changing denominators remain implicit. Notes should state exclusions or analysis populations needed to interpret percentages and estimates.
Should uncertainty appear in every figure?
Show uncertainty when it is relevant to judging estimates or comparisons, using an appropriate interval or distribution representation. The display and legend should identify what the interval means. Do not add error bars decoratively or omit them to make differences look decisive.
Assessment move
For each display, write one sentence describing its job. Verify data values against the analysis, then check titles, labels, units, counts, denominators, notes and abbreviations. Inspect whether the scale or geometry could distort comparison and whether groups remain distinguishable without colour.
In the accompanying prose, report the estimate, direction and uncertainty before threshold language and avoid duplicating full rows. Search the complete manuscript for each sample size and central value so the abstract, results and displays remain consistent. Use a display audit sheet with separate rows for factual accuracy, analytical appropriateness, visual integrity and accessibility.
Recalculate several cells from the underlying output and confirm that percentages use the stated denominator. Read the title and notes without the main text, then view the display in grayscale. Finish by asking what a hurried reader might conclude and whether the design makes that inference warranted.
