git clone the repo locally

Install Anaconda

Install UMAP:

conda install umap-learn -c conda-forge

Verify that the environment defaults to python 3

npm install

sh setup-develop.sh
sh update-develop.sh

Note: We recommend using Jupyter in full screen mode when running VAINE. Widget may appear truncated otherwise.

Open Jupyter Notebook:

jupyter notebook

Navigate to notebooks/VAINE demo

This file should run without errors

Load dataset from local filepath or url:

import pandas as pd
df = pd.read_json(url)
df = pd.read_csv(url)

If necessary, create an index column. Note that indices need to be unique:

df = df.set_index('variable')

Define treatment, outcome and ignore variables. VAINE currently only handles numerical variables, categorical variables should be ignored.

treatments = df.columns[df.columns.str.startswith('Treatment')].tolist()
outcomes = df.columns[df.columns.str.startswith('Outcome')].tolist()

ignore = ['variable 1', 'variable 2']

Run VAINE:

vaineWidget(df, treatments, outcomes, ignore)