Significance tests

[strengejacke]

Not sure if you think this fits into effectsize, but I'm missing convenient wrappers around common significance tests, including options to use weights. I have done this for now in sjstats:
https://strengejacke.github.io/sjstats/reference/index.html#summary-statistics-and-tests (e.g. https://strengejacke.github.io/sjstats/reference/chi_squared_test.html).

Let me know what you think. The functions are more or less in easystats style, so would be easy to move over here.

[mattansb]

I would love to have weights in various effect sizes (#388 ) - do you mean to have a function like t_test() etc?

[bwiernik]

What sort of weights are you thinking about? If we mean survey weights used to adjust for nonresponse or unrepresentative sampling, then we need to ensure that standard errors and significance tests are adjusted appropriately for the weighting design effects (cf. the survey package)

Sorry, I edited your comment instead of the quoted one 🤣

[strengejacke]

I would love to have weights in various effect sizes (#388 ) - do you mean to have a function like t_test() etc?

Yes

[strengejacke]

What sort of weights are you thinking about?

Simple frequency weights.

If we mean survey weights used to adjust for nonresponse or unrepresentative sampling, then we need to ensure that standard errors and significance tests are adjusted appropriately for the weighting design effects (cf. the survey package)

I use this function (or the package) for "simple" weights already, but I think it should, for now, be limited to simple frequency weights.

These kinds of weights are commonly used in my field

[strengejacke]

bump

What's your opinion about having these functions in effectsize?

[mattansb]

I'm not sure... It feels out of scope (?) but I can't think of a better place for these in {easystats} (if you're idea was to port those from your sj-verse?)

We can have those here, but I think it should be accompanied by having weighting options for the corresponding effect sizes as well.

[strengejacke]

Ok, make sense. Weighting options make sense in general, even if we don't port the tests from sj to easystats.