How to report P values in journals.
When you write up your results, how should you report P values?
Don't overemphasize P values
A few pointers:
- Consider emphsizing the effect size and confidence interval, rather than a P value. The effect size can be a difference or a ratio or a correlation coefficient (or something else) and all can be reported with a 95% confidence interval to demonstrate how precisely you have determined the effect size. This is more useful, in many cases, than P values.
- Don't just say if the P value is greater or less than 0.05 (or some other value). If you can, give the P value as a number.
- With multiple comparisons, it is not possible to give individual P values for each comparison, but it is possible in many cases to report multiplicity adjusted P values.
- If you computed many P values, show them all. At a mimunum, state the number of comparisons you made. Performing many analyses on your data and then just reporting the comparisons that result in a small P value will create misleading results.
- It doesn't help to report a P value unless you clearly state what test was used to compute it.
Many people add asterisks to tables and graphs to show how low the P value is. The standards for one to three asterisks are quite standard (<0.05, <0.01, <0.001), and both the NEJM and APA agree. Prism (since 5.04/d) will also show four asterisks when the P value is less than 0.0001.
|GP Prism up to 5.04/d||*||**||***||***|
|GP Prism 5.04/d and later||*||**||***||****|
Make sure it is clear what null hypothesis the P value is testing
Every P value tests a null hypothesis, so your readers need to be sure what the P value is testing. If you put an asterisk on a graph, ti has to be clear exactly what comparison was made.
Issues that copy editiors care about (but scientists don't)
How many digits to report? Leading zero or not?
How exactly should P values be reported? Three styles are summarized below:
The Americal Psychological Association (APA) has published an extensive style guide used for publishing in the social sciences. One rule is to not include the zero before the decimal point when reporitng a P value: "There are some values that by definition can never exceed 1.0. The omission of the leading zero is a visual indicator of this restricted range. The most common cases are p values and correlations.". They also state to report two or three digits after the decimal point. The P value in the first column of the table is shown as ".123" in APA style, but you could read their style to mean ".12".
The New England Journal of Medicine (NEJM) states: "Except when one-sided tests are required by study design, such as in noninferiority trials, all reported P values should be two-sided. In general, P values larger than 0.01 should be reported to two decimal places, those between 0.01 and 0.001 to three decimal places; P values smaller than 0.001 should be reported as P<0.001. Notable exceptions to this policy include P values arising in the application of stopping rules to the analysis of clinical trials and genetic-screening studies."
We never intended to create a style, but GraphPad (GP) programs are in wide use, so many people follow our lead. GraphPad InStat and Prism always report a zero before the decimal point, and four digits after. If the P value is less than 0.0001, we report "<0.0001".
"P value" or "p value"
There is no uniform style.
The APA suggest "p value" The p is lowercase and italicized, and there is no hyphen between "p" and "value".
GraphPad has adapted the style "P value", which is used by the NEJM and journals. The P is upper case and not italicized, and there is no hyphen between "P" and "value".
Sometimes, you see "p-value". Note the hyphen.