Revise the way we calculate % in ROPE (SGPV) for frequentist models? #849

[strengejacke]

#849

[bwiernik]

My preference for both freq and Bayesian would be to default to computing as the proportion of the posterior that's in the ROPE (so rope_ci = 1). That would make the % be 1 - the equivalence test p value IIRC

[strengejacke]

I think this doesn't work right now. We have this code, with hardcoded 3.28:

  x <- bayestestR::distribution_normal(
    n = 1000,
    mean = ci_range[2] - (diff_ci / 2),
    sd = diff_ci / 3.28
  )

I guess this works approximately only for 95% CI, not for others...