The following guidelines for the reporting of statisical analysis of data were phased in at JARO, beginning in April, 2012.
Statistical results should be reported following these guidelines:
In the Methods, state how the data are reported in the manuscript (standard deviation or standard error of the mean), the rationale for the choice of specific tests or uncommon tests, and the software that was used to compute statistical results.
Statements that indicate the presence of “significant effects” should be supported by an appropriate statistical test.
Use a statistical test that is appropriate for the data. Usually this involves first determining if the datasets are drawn from a Gaussian distribution (in which case parametric tests can be used), and determining whether the variances are equal (or homoscedastic). Use a nonparametric or permutation type test if these critieria are not met.
State the type of test that was used, including the degrees of freedom, the resulting test value, and the exact P-value (to 2 significant figures) that the result occurred at chance under the null hypothesis. For t-tests, always state whether the test is for a one-sided or two-sided hypothesis.
Format: The format of the description of the statistical results should indicate the degrees of freedom, the statistic value, and the P value, as in these examples:
|F(3,21) = 5.62, P = 0.0054|
|t(7) = 4.582, P = 0.0025|
|r2(9) = 0.77, P = 0.0004|
To avoid ambiguities, all statistical variables should be italicized (F, t, P).
The type of post-hoc tests used when following any ANOVA with multiple comparisons should be identified.
If you have a large number of statistical tests, you may wish to include those results in a summary table to make the manuscript easier to read. Please refer to the table in the text.
Reporting P values with inequalities (“p < 0.05”) should be limited to figures, or for post-hoc tests (multiple comparisons) if no exact value is reported by the software.