How to report statistics in a class report or manuscript.md

How to report statistics in a class report or manuscript

This is a work in progress…feel free to suggest

Relevant to : ecology, evolution, behaviour

• Provide all the information required (with the exception of data, which should be linked to in a manuscript) to replicate the test in the "Methods". This may include:
  • Description of the model (response variable, explanatory variable, structure)
  • Reference for more exotic tests or for the application of standard tests in a specific context
  • Any data manipulation (transformation, dropped data)
  • Statistical package used where relevant, including references for add on packages where these are essential to the analysis
• Support all statements in "results" section:
  • Example structures:
    • Statement (mean [95% confidence interval), difference of xx [xx-xx], ANOVA, Factor A, F_{numerator_degrees_of_freedom,denominator_degrees_of_freedom} =xx, p=xx)
    • Values in level A were on average xx [xx-xx] bigger than values in level B (test, test value with degrees of freedom, p-value).
  • When there is a significant difference or trend with a biologically relevant parameter estimate and a measure of its variance (eg. mean difference with 95% confidence interval or slope of a regression). It does not make sense to report differences/trends when these could easily be obtained by chance (i.e. not statistically significant)
  • Where relevant, either an inline report of the statistical test or reference to a table containing these test results.
    • Test results should include:
      • The name of the test
      • The variable being tested (if multiple possibilities)
      • Values required to calculate the exact p-value
        • The exact value of the test results (F statistic)
        • The degrees of freedom
      • The exact p-value

Rationales for reporting all exact statistical test values, even when tests are not significant:

• Allows reader/reviewer to have confidence in provided values by allowing them to compare values to their understanding of the underlying test (eg. do the reporting effect size match the level of significance?); to their recalculated p-values based on provided stats (eg. are there any typos?), or even to their recalculations based on the original data ("reproducible research").
• Allows readers to extract statistics for meta-analysis

References:

A useful, if incomplete guide:

https://onlinelibrary.wiley.com/doi/pdf/10.1111/j.1095-8649.2011.02914.x

Please comment, suggest edits (push), etc. Is there a good guide somewhere else?