Skip to main content

calibrated prediction across diverse contexts

Project description

CalPred (Calibrated Prediction Intervals for Polygenic Scores Across Diverse Contexts)

See companion manuscript github repository for analysis scripts used in the manuscript.

Installation

# calpred calls R packages statmod and Rchoice in fitting the model
Rscript -e "install.packages(c('statmod', 'Rchoice'), repos='https://cran.rstudio.com')"
pip install calpred

Quick example

import numpy as np
import pandas as pd
import matplotlib.pyplot as plt
import calpred

np.random.seed(42)
n = 1000
pgs = np.random.normal(size=n)
age = np.random.normal(loc=40, scale=10, size=n)
sex = np.random.binomial(n=1, p=0.5, size=n)

y_mean = 8 + pgs * 0.5 + age * -0.2 + sex * 0.5
y_sd = np.sqrt(np.exp(2 + age * -0.03 + sex * 1))
y = np.random.normal(loc=y_mean, scale=y_sd)

df = pd.DataFrame({"intercept": 1, "pgs": pgs, "age": age, "sex": sex, "y": y})

# x and z are the columns for fitting the mean and standard deviation
x = z = df[["intercept", "pgs", "age", "sex"]]
model = calpred.fit(y=df["y"], x=x, z=z)

# prediction mean and [low, high] for 90% prediction interval
df["pred_mean"], df["pred_sd"] = calpred.predict(x=x, z=z, model_fit=model)
df["pred_low"] = df["pred_mean"] - df["pred_sd"] * 1.645
df["pred_high"] = df["pred_mean"] + df["pred_sd"] * 1.645


# show prediction intervals at 5% / 95% quantile of prediction mean
fig, ax = plt.subplots(figsize=(4, 4), dpi=150)
ax.scatter(df["pred_mean"], df["y"], s=4)
ax.axline((0, 0), slope=1, ls="--", color="red")

idx1 = df.sort_values("pred_mean").index[int(n * 0.05)]
idx2 = df.sort_values("pred_mean").index[int(n * 0.95)]

for idx in [idx1, idx2]:
    ax.errorbar(
        x=df.loc[idx, "pred_mean"],
        y=df.loc[idx, "pred_mean"],
        yerr=df.loc[idx, "pred_sd"] * 1.645,
        color="red",
        capsize=3,
        lw=1,
    )
fig.show()

Upload to PyPI (for developers)

python setup.py sdist
twine upload dist/*

Project details


Download files

Download the file for your platform. If you're not sure which to choose, learn more about installing packages.

Source Distribution

calpred-0.1.1.tar.gz (16.8 kB view details)

Uploaded Source

File details

Details for the file calpred-0.1.1.tar.gz.

File metadata

  • Download URL: calpred-0.1.1.tar.gz
  • Upload date:
  • Size: 16.8 kB
  • Tags: Source
  • Uploaded using Trusted Publishing? No
  • Uploaded via: twine/4.0.2 CPython/3.9.13

File hashes

Hashes for calpred-0.1.1.tar.gz
Algorithm Hash digest
SHA256 ce5357c615d6d62669055df74b88b3aedbba7c90184e07883fe573b96f36cda1
MD5 e614e8fd8c8257fe76b01b582a51c70d
BLAKE2b-256 f87b1d7b11038ae178c382c547af111cfec326e5f13785e575a0adff773e22a8

See more details on using hashes here.

Supported by

AWS AWS Cloud computing and Security Sponsor Datadog Datadog Monitoring Fastly Fastly CDN Google Google Download Analytics Microsoft Microsoft PSF Sponsor Pingdom Pingdom Monitoring Sentry Sentry Error logging StatusPage StatusPage Status page