The p2prev
package provides a simplified interface to fit
from p2prev import PCurveMixture # import model class from p2prev
import numpy as np
pvals = [0.00060, 0.02999, 0.04939, 0.94601]
pvals = np.array(pvals) # p-vals should be in an array
model = PCurveMixture(pvals) # feed model the data
model.fit() # fit the model parameters to the data
model.summary() # prints results
PCurveMixture
results can be compared between two independent groups to estimate between-group differences in effect prevalence or in within-subject effect size (given that the subject shows the effect), for which we provide an example. We also provide a p2prev.PCurveWithinGroupDifference
class for estimating the prevalence difference between two tests performed on the same group of subjects (or e.g. the same group of subjects is given the same test in two experimental conditions).
Check out our tutorial examples and Documentation for details.
We recommend installing pymc
before trying to install p2prev
. (I recommend version 5.0 or greater, though I think p2prev
will run with Version 4.) Installing p2prev
as below will attempt to install pymc
using Pip, but I find that pymc
is much more likely to install correctly using Anaconda as suggested in the PyMC documentation.
Once you've installed pymc
, you can simply run
pip install p2prev
to install the latest stable version.
You could also install the development version with
pip install git+https://github.com/john-veillette/p2prev.git
but do so at your own peril! I may sometimes break things as I work on the development version.
If you use p2prev
package, please be so kind as to cite our preprint that describes and validates the method.
We have also minted a DOI so that you can cite the p2prev
package itself (or other contents of this repository) and be confident that the link in your citation will always lead to the same place. The DOI for the current version should always be near the top of this page on GitHub, or you can use the permanent DOI (10.5281/zenodo.11459064), which always links to the latest release.
Importantly, p2prev
is only a wrapper around a model implemented with the excellent PyMC package, which performs the hardest parts of model fitting for us. So if you use p2prev
, please also cite PyMC so its developers can get credit for their work.