Andrew Vigotsky writes:
Now that abandoning significance and embracing uncertainty is in the air, we think this package, which runs in R or Stata, may be of interest to both you and your readers.
Concurve plots consonance curves, p-value functions, and S-value functions to allow readers and researchers to get a better feel of the range of values with which their data are compatible. The package is able to do this for everything from mean differences to meta-analytic estimates.
These confidence intervals aren’t really my thing—I prefer a Bayesian approach—but I’m sharing this as it might interest some of you.
For a frequentist, a consonance (p-value) function may be one of the best tools available. I think D. A. S. Fraser summarized it nicely in his recent TAS paper,
“The full p-value function arguably records the full measurement information that a user should be entitled to know! We also note that the p-value function is widely available for any scalar interest parameter and any model-data combination with just moderate regularity, and has available computational procedures.
In addition to the overt statistical position, the p-value function also provides easily and accurately many of the familiar types of summary information: a median estimate of the parameter; a one-sided test statistic for a scalar parameter value at any chosen level; the related power function; a lower confidence bound at any level; an upper confidence bound at any level; and confidence intervals with chosen upper and lower confidence limits. The p value reports all the common inference material, but with high accuracy, basic uniqueness, and wide generality.
From a scientific perspective, the likelihood function and p-value function provide the basis for scientific judgments by an investigator, and by other investigators who might have interest. It thus replaces a blunt yes or no decision by an opportunity for appropriate informed judgment.”
https://doi.org/10.1080/00031305.2018.1556735
I’ve wanted something like this for a while, nice!
Actually, I struggle to imagine issues with ‘standard practice’ frequentist inference that these don’t go a long way to address…can anyone think of a case where these don’t give more reasonable, intuitive results than current practice?