You can deploy this model and start earning money today!

snowfire/
wellspring-latentwb1

Name: wellspring-latentwb1
Availability: InStock

Views

No views yet

WellSpring — LatentWB-1

A transparent latent well-being index model. It distils six intertwined national well-being drivers into two independent, human-readable indices and predicts a country's well-being (Cantril-ladder) score from them:

Development Index — GDP per capita, healthy life expectancy, social support, freedom
Social / Civic Virtue Index — generosity, clean/trustworthy institutions

well-being = 5.407 + 1.132 · DEV + 0.119 · SOC

Performance

R² = 0.78 on the 2019 World Happiness Report cross-section (156 countries).
Matches a conventional six-driver linear model on fit, and beats it on out-of-sample accuracy and coefficient stability while using only two inputs.

License

Apache-2.0 — free to use, modify, and redistribute with attribution.

Files

index_specification.json — standardization parameters + index definitions/weights
latentwb1/regression.json — fitted regression (coefficients, CIs, fit stats)

Usage

python

import json
from huggingface_hub import hf_hub_download

spec = json.load(open(hf_hub_download("snowfire/wellspring-latentwb1", "index_specification.json")))
reg  = json.load(open(hf_hub_download("snowfire/wellspring-latentwb1", "latentwb1/regression.json")))
p = reg["fit"]["params"]  # {const, DEV_index_std, SOC_index_std}

def predict(row):  # row: dict of the six raw drivers
    z = {d: (row[d] - spec["standardize_mean"][d]) / spec["standardize_std"][d] for d in spec["drivers"]}
    idx = {code: sum(ax["weights"][m] * z[m] for m in ax["members"]) for code, ax in spec["axes"].items()}
    return p["const"] + p["DEV_index_std"] * idx["DEV"] + p["SOC_index_std"] * idx["SOC"]

Data

Developed on the World Happiness Report (2015–2019), public domain (CC0), from the Gallup World Poll.