Skip to content

The specification of permutations in anova.cca() has issues, and the number of permutations should be 999, not 1000. #17

Description

@pedrohbraga

The definition of the argument to change permutations in the presentations and in the books is done as in anova.cca(... , step = 1000).

step is not an argument included in or used by anova.cca(). Somehow, anova.cca() does not throw errors if one uses gibberish arguments, but the argument to change the permutations is permutations, as in anova.cca(... , permutations = how(nperms = 999).

How how() works here is less trivial than just changing the number attributed to nperms and needs to be reviewed in this workshop. See this document for help.

Moreover, in certain slides, the number of permutations used is 1,000 instead of 999. There is no explanation provided for this choice.

The p-value for permutations in the models there is calculated as p = (N + 1) / (n + 1), where n is the number of permutations requested, and N is the number of values of the test statistics (any type) that have been as extreme as the test statistic for the observed data.

That addition of +1 is what explains why we often see 999 or 199 permutations instead of 1000 or 200.

The reason that a division of 90/(99 +1) is often preferred over a 90/(100 + 1) is that it makes p-values "pretty" or easily interpretable in terms of the number of test statistics from permuted values that came as extreme as the observed test statistic.

Metadata

Metadata

Assignees

No one assigned

    Labels

    bugSomething isn't working

    Type

    No type

    Projects

    No projects

    Milestone

    No milestone

    Relationships

    None yet

    Development

    No branches or pull requests

    Issue actions