Installation¶

Requirements¶

Python 3.14.7 or later, below 3.15.
No R installation. The reference numbers from R are committed as fixtures and are only needed by the test suite.

metafopy depends on NumPy, SciPy, pandas, formulaic (model formulas such as '~ ablat') and matplotlib (plots). They are installed with it.

Install¶

metafopy is pre-release and is not on PyPI yet. Install it from GitHub with pip or uv:

pip install "metafopy @ git+https://github.com/bnomis/metafopy"
# or
uv add "metafopy @ git+https://github.com/bnomis/metafopy"

To work on metafopy itself, clone the repository and follow the development section of the README.

Interactive charts (optional)¶

metafopy.interactive builds interactive forest and funnel charts with Altair (see Plots), and the showcase notebook links them. Install the interactive extra to use them (it adds Altair and vl-convert, which saves the charts as SVG or PNG):

pip install "metafopy[interactive] @ git+https://github.com/bnomis/metafopy"

Results also render as HTML tables in Jupyter, which needs no extra.

Check the installation¶

import metafopy

print(metafopy.__version__)

0.0.1

Then fit a first model on a bundled dataset:

from metafopy import datasets, rma

dat = datasets.load('dat.hackshaw1998')
print(rma(data=dat))

Random-Effects Model (k = 37; tau^2 estimator: REML)

tau^2 (estimated amount of total heterogeneity): 0.0224 (SE = 0.0178)
tau (square root of estimated tau^2 value):      0.1497
I^2 (total heterogeneity / total variability):   29.57%
H^2 (total variability / sampling variability):  1.42

Test for Heterogeneity:
Q(df = 36) = 47.4979, p-val = 0.0952

Model Results:

estimate      se    zval    pval   ci.lb   ci.ub
  0.2189  0.0494  4.4313  <.0001  0.1221  0.3157  ***

---
Signif. codes:  0 '***' 0.001 '**' 0.01 '*' 0.05 '.' 0.1 ' ' 1

If this prints a "Random-Effects Model" summary, the install works. dat.hackshaw1998 holds the studies of Hackshaw, Law and Wald (1997) on lung cancer and passive smoking, https://doi.org/10.1136/bmj.315.7114.980.

Plots without a display¶

metafopy draws with matplotlib. On a server or in CI with no display, select a non-interactive backend before plotting:

import matplotlib

matplotlib.use('Agg')