Table Presentation Guidelines
Compiled by James Beard
Introduction
The tables and figures are the most important elements of a research article, and should enable the reader to understand the main findings of the study independently of the text.
Research articles will typically include several tables. These will detail (at least):
- the characteristics of the study population
- the main results
- secondary findings
Although the study population will often be described in a single table, multiple tables may be necessary or desirable for the main results and secondary findings.
If a table has been previously published elsewhere you must obtain the consent of the copyright holder to republish it (unless the copyright notice allows non-commercial re-use) and the source must be acknowledged.
Formatting and content of tables
Each table must have a number and an informative title, and be mentioned in, and discussed in, the text. The numbering should be sequential in the order they are referred to in the text, Table 1, Table 2, etcetera.
Keep tables simple, and do not include irrelevant data, especially data that are not referred to in the text. Very large tables will generally not be accepted for publication in their entirety, but can be offered as an addendum to the publication, available on request from the corresponding author. Do not present the same information in both tables and figures.
Tables should be inserted into the manuscript as editable cell-based tables, not as images or constructed with text boxes or tabs. Do not use shading. Each row and column of the table must be a separate element of the table - do not use the Enter key to create new rows or spacing / tabs to create new columns.
Table titles may include an indication of the sample size, for example (N=81). But often it will also be necessary to include a total row at the bottom of tables, and sometimes a total column.
Row and column labels should be clear but concise, whilst avoiding abbreviations if possible, and should include units where appropriate.
Raw numbers and percentages should be in the same table column labelled n (%), not N (%).
Numbers and row labels should be left-justified in cells. In cells containing a number and a percentage, there should be a space between the number and the (.
Percentages should be presented with the same number of decimal places (normally one) for all cells in a table. That is, (15.6), (10.0), (100.0), not (15.6), (10), (100). Do not include the % sign, as this will have been included in the column heading.
Columns of probabilities should be labelled p-value, not P-value, P or p. They should be displayed to 3 decimal places, so 0.020, not 0.02. Values rounded to zero should be reported as <0.001.
Confidence intervals should be displayed as (lower limit, upper limit), for example (0.72, 1.27). They may be presented in the same column as numbers, in which case, a space should precede the ( or in a separate column immediately to the right of the numbers column. The label for the column should include the significance level, for example, 95% CI.
Odds, risk or hazard ratios should normally be presented to two decimal places. But if all the ratios presented within a single table have large confidence intervals, it is preferable to use one decimal place. For example, 1.2 (0.1, 9.7) is more appropriate than 1.23 (0.13, 9.68).
There is almost never justification for presenting numbers with more than 3 significant digits.
When reporting chi-squared tests, it is rarely useful to include the actual value of the chi-squared statistic, the probability is enough.
If the chi-squared statistic is reported, it is usually best not to use the actual Greek letter chi (χ) in the label (or in a table title), as it may look like the Western European letter X in the fonts used in the production version of the article.
Category labels for a variable should be exhaustive and non-overlapping. This is easy to get wrong for numerical ranges – for age (years) for example, the categories might be <20, 20–24, 25–29, ≥30, not <20, 20–24, 25–29, >30 (30 omitted) or <20, 20–25, 25–30, >30 (25 in two categories).
Category labels that are numeric ranges should be lower to upper, even when the quantities are negative, so -3 to <-2, not <-2 to -3.
Consider grouping categories when one or more contain very small numbers. For example, education level may have been collected with categories None, Primary, Secondary, University, but only 5 people out of a sample of 400 had university education - in this case it would usually be sensible to combine the Secondary and University categories for reporting. This is particularly important if statistical tests are reported in the table.
Table footnotes should be indicated with consecutive use of the following symbols: * † ‡ § ¶ ‖ then ** †† ‡‡ etcetera. Abbreviations used in a table, but not elsewhere in the article text, should be defined underneath the table. It is not necessary to define abbreviations found in nearly all tables, such as CI, SD, IQR (Confidence Interval, Standard Deviation, Interquartile Range).
It is often easier to prepare a table in Microsoft Excel, rather than in Word, and then copy and paste the table into the article Word document. This is because it is easier to control the formatting of the table, it is possible to do calculations within the table, and cells can be created by copying or concatenating information from other cells. Indeed, the example tables below were created using Excel. When pasting a table from Excel into Word, paste it as editable text, not as an Excel object, as a picture or as a link.
Example tables
All the following use invented data. Table 1 is an (rather short) example of a table showing participant characteristics. Note related characteristics grouped together. Table 2 shows an analysis of behaviour of interest by sex and socio-economic status, with chi-squared test probabilities. Note the use of rows, rather than a column (as in Table 1) for the variable labels, to save horizontal space. Table 3 shows the output of a logistic regression analysis investigating the relationship of acceptance into post-graduate studies (the dependent variable) to Graduate Record Exam score, Grade Point Average and the prestige of the student's undergraduate institution (the independent variables). The table shows both the crude, unadjusted, odds ratio (cOR) and the adjusted odds ratio (aOR). Note that the percentages in the Total column are column percentages, whilst those in the other columns are row percentages.
Table 1. Patient characteristics (N=162)
|
n (%) |
|
Age (years) |
< 20 |
18 (11.1) |
20 – 24 |
52 (32.1) |
|
25 – 29 |
41 (25.3) |
|
≥ 30 |
51 (31.5) |
|
Gestational age (weeks) |
< 10 |
46 (28.4) |
≥ 10 |
94 (58.0) |
|
Missing |
22 (13.6) |
|
Gravidity |
1 |
75 (46.2) |
2 |
46 (28.4) |
|
≥ 3 |
41 (25.3) |
|
Marital status |
Single |
32 (19.8) |
Married |
105 (64.8) |
|
Divorced |
25 (15.4) |
|
Highest educational level |
None |
31 (19.1) |
Primary |
96 (59.3) |
|
Secondary or higher |
35 (21.6) |
Table 2. Description of starting high school students' choices, according to sex and socio-economic status (N=200)
|
Sex |
Socio-economic status |
|
|||||
Male n (%) |
Female n (%) |
p-value |
Low n (%) |
Middle n (%) |
High n (%) |
p-value |
Total n (%) |
|
Program Type |
|
|
|
|
|
|
|
|
General |
21 (23.1) |
24 (22.0) |
0.974 |
16 (34.0) |
20 (21.1) |
9 (15.5) |
0.002 |
45 (22.5) |
Academic |
47 (51.7) |
58 (53.2) |
|
19 (40.4) |
44 (46.3) |
42 (72.4) |
|
105 (52.5) |
Vocation |
23 (25.3) |
27 (24.8) |
|
12 (25.5) |
31 (32.6) |
7 (12.1) |
|
50 (25.0) |
School Type |
|
|
|
|
|
|
|
|
Public |
77 (84.6) |
91 (83.5) |
0.828 |
45 (95.7) |
76 (80.0) |
47 (81.0) |
0.030* |
168 (84.0) |
Private |
14 (15.4) |
18 (16.5) |
|
2 (4.3) |
19 (20.0) |
11 (19.0) |
|
32 (16.0) |
Total |
91 (100.0) |
109 (100.0) |
|
47 (100.0) |
95 (100.0) |
58 (100.0) |
|
200 (100.0) |
* Fisher’s exact test used
Source: https://stats.idre.ucla.edu/stat/data/hsbdemo.dta
Table 3. Logistic regression exploring relationship between student characteristics and admission to post-graduate study
|
Total |
Not admitted |
Admitted |
cOR (95% CI) |
p-value |
aOR (95% CI) |
p-value |
Graduate record exam score |
|
|
|
||||
<580 |
174 (43.5) |
131 (75.3) |
43 (24.7) |
1.00 |
|
1.00 |
|
≥580 |
226 (56.5) |
142 (62.8) |
84 (37.2) |
1.80 (1.16, 2.79) |
0.008 |
1.31 (0.82, 2.10) |
0.261 |
Grade point average |
|
|
|
||||
<3.395 |
200 (50.0) |
155 (77.5) |
45 (22.5) |
1.00 |
|
1.00 |
|
≥3.395 |
200 (50.0) |
118 (59.0) |
82 (41.0) |
2.39 (1.55, 3.70) |
<0.001 |
2.33 (1.46, 3.70) |
<0.001 |
Prestige of undergraduate institution |
|
|
|
||||
Lower |
188 (47.0) |
148 (78.7) |
40 (21.3) |
1.00 |
|
1.00 |
|
Higher |
212 (53.0) |
125 (59.0) |
87 (41.0) |
2.58 (1.65, 4.01) |
<0.001 |
2.61 (1.65, 4.13) |
<0.001 |
Total |
400 (100.0) |
273 (68.3) |
127 (31.8) |
|
|
|
|
Source: https://stats.idre.ucla.edu/stat/stata/dae/binary.dta
Acknowledgements
Some of the text above was drawn from:
Appendix 23.1: Guidance on how to write a scientific paper in Smith PG, Morrow RH, Ross DA (Ed.). Field Trials of Health Interventions: A Toolbox: 3rd ed. Oxford University Press, Oxford; 2015 https://doi.org/10.1093/med/9780198732860.001.0001
and
South African Medical Journal. Author Guidelines http://www.samj.org.za/index.php/samj/about/submissions#Tables
Table 2 uses invented data from UCLA Statistical Consulting Group. Multinomial Regression, Stata data analysis examples https://stats.oarc.ucla.edu/stata/dae/multinomiallogistic-regression/
Table 3 uses the invented data and example at UCLA Statistical Consulting Group. Logistic Regression, Stata data analysis examples https://stats.oarc.ucla.edu/stata/dae/logistic-regression/