Writing-up statistical results is as important as selecting the correct statistical test to apply to your data. To convey the message to the readers, statistics need to be reported clearly and in a comprehensible way. Individual academic journals may have slightly different formats and guidelines for how results should be reported. To facilitate the review and editorial process for publication in the British Journal Oral and Maxillofacial Surgery (BJOMS) – and bearing in mind that most submissions will undergo a statistical review alongside the technical review – these guidelines are intended to assist authors in their endeavours to publish in BJOMS (and should be read in conjunction with Evans et al ):
The reporting of statistical results should follow the sequence:
Description of patients studied: age, number (percentage) of males/females, clinical descriptors, and sociodemographic details. These should all be summarised in table(s);
Results of the primary research question, including tables and figures, then secondary aims, and so on.
Means and standard deviations (except for age, see below*) should be presented for continuous variables (that is, data that can be broken down into increasingly smaller units, such as age or height).
For categorical or discrete data (including nominal or ordinal data, that is, data that can be counted, for example, number of tumour sites, type and stage of tumour), median and ranges should be used, for instance minimum to maximum value: median 6 (range = 5 – 7).
Percentages (including the number of X/total N) or proportions, or both, should be included for categorical data.
Some further examples:
Mean score = 5.67 (SD ‡ 2.25)
*Mean age = 45.2 (range: 21.5 – 77.8)
Percentage edentulous patients = 32% (48/150)
‡ Standard deviation. This (and other abbreviations) should always be spelled out in full when first referenced.
When reporting means also include the 95% confidence intervals (CI).
Include the test value and degrees of freedom when reporting statistical results. The general format for reporting these is: test value (degree(s) of freedom) = test value, p = p-value. Some examples of these and how to report them are provided below:
T-test:t(2) = 2.58, p = 0.01
One-way ANOVA † : F(4) = 1.38, p = 0.18
Two-way ANOVA: F(2,8) = 7.13, p = 0.021
Chi-squared test: χ 2 (5) = 8.34, p = 0.14
The results of correlations and non-parametric tests should just include the test statistic (for example, r, ρ or rho, z, and so on.) and the exact p values.
† Analysis of variance.
Test and p values may be included in either the main body of the text or tables. This will depend on context: when highlighting a specific result in the text then report the result as described in point 4. When reporting a series of results, individual p values can be used for each result in the body of the text with the details provided in the associated table.
Exact p values should always be reported, for example, p = 0.021, unless the p value is less than 0.001, in which case this may be reported as p < 0.001. Exact p values should always be reported even when the result is not significant:
Do not report p values as p < 0.05, p < 0.01, or p > 0.05. Do not include the abbreviation “NS” (“not significant” or “non-significant”) when reporting results. If the results are not significant simply state this.
Some statistical packages report p values less than 0.001, as 0.000: in cases such as these the p value should be reported as p < 0.001 (and not 0.000).
To avoid confusion, do not use the word significant alone when reporting results, for example, a significant difference, unless the result is statistically significant . The journal uses the word “significant(ly)” only in a statistical sense, so there is no need to add any further qualification to this, such as highly or very . It is sufficient to simply report that the results reached significance.
A picture paints a thousand words: include graphs appropriate to the results you are describing but avoid pie charts.
Finally, for more complex statistics, for instance, when reporting the results of regressions, include R 2 or the variance explained by the regression model. The reason for this inclusion is that significant predictors may not be particularly meaningful, if the model explains little or none of the variance. In addition to this the beta coefficients (standardised or unstandardised) should be reported along with the t-statistic and exact p values for each coefficient.
You may also need