Datasets

metafopy ships the datasets of the R package metadat that the metafor examples use. Load one by its R name, with or without the dat. prefix, as a pandas.DataFrame:

from metafopy import datasets

print(datasets.list_datasets()[:5])
dat = datasets.load('dat.bcg')
print(dat.head())
print(datasets.load('bcg').equals(dat))

['dat.assink2016', 'dat.bangertdrowns2004', 'dat.bcg', 'dat.berkey1998', 'dat.cohen1981']
   trial                author  year  tpos  ...  cpos   cneg  ablat      alloc
0      1               Aronson  1948     4  ...    11    128     44     random
1      2      Ferguson & Simes  1949     6  ...    29    274     55     random
2      3       Rosenthal et al  1960     3  ...    11    209     42     random
3      4     Hart & Sutherland  1977    62  ...   248  12619     52     random
4      5  Frimodt-Moller et al  1973    33  ...    47   5761     13  alternate

[5 rows x 9 columns]
True

list_datasets() lists every bundled dataset. Each load returns a fresh copy that you can modify.

The metadat documentation describes the variables and cites the original source of each dataset: https://wviechtb.github.io/metadat/. Columns keep their R names, including dots (or.lb), so use brackets (dat['or.lb']) to access them.

Trees

dat.moura2021 comes with a phylogenetic tree, loaded with datasets.load_tree('dat.moura2021.tree'). The result is a PhyloTree with the tip labels (tip_labels) and the variance-covariance or correlation matrix of the tips (vcv), which is what rma_mv needs for phylogenetic models. See Multilevel and multivariate models.

Using your own data

Any DataFrame works. Give column names as strings with data=, or pass arrays directly. Missing values are handled as in metafor: na_action='omit' (the default) drops studies with missing values, 'fail' raises an error.