sturdy-stats-sdk 1.0.50

SDK for the Sturdy Statistics API

Sturdy Stats SDK

This is the sdk for the Sturdy Statistics API. We host a series of public indicies trained on Earnings Calls, ArXiv, HackerNews, and various news streams that anyone can use for public data analysis. Uploading data requires signing up at https://sturdystatistics.com in order to create an api key.

Installation

Core API: pip install sturdy-stats-sdk

Regression Extension: pip install sturdy-stats-sdk[regression]

Resources

Explore our gallery to browse visualization created by the sturdy-stats-sdk. Follow along with our quickstart to hit the ground running or browse our advanced examples to perform rigorous analyses.

Technical Features

Automatic Structuring of Unstructured Text Data

Convert unstructured documents into structured formats, allowing seamless analysis alongside traditional tabular data. Learn More >

Explainable Text Classification

Gain clear insights into how text data is categorized, while enhancing transparency and trust in your analyses. Learn More >

Effective with Small Datasets

Achieve meaningful results even with limited data, making our solutions accessible to organizations of all sizes. Learn More >

Powerful Search Capabilities

Leverage our robust search API to retrieve and analyze specific information within your unstructured data. Learn More >

Comprehensive Data Lake

Store and analyze all your data — structured and unstructured — in one place, facilitating holistic insights. Learn More >

Quickstart

Explore Your Data

from sturdystats import Index, Job
import plotly.express as px

index = Index(id="index_99051ff1489844878fd792784d7baa90")
topic_df = index.topicSearch()
fig = px.sunburst(
    topic_df, 
    path=["topic_group_short_title", "short_title"],
    values="prevalence", 
    hover_data=["topic_id"]
)

Run SQL queries against your unstructured ata

topic_id = 12
df = pd.DataFrame(index.queryMeta(f"""
SELECT
    quarter,
    sum(sparse_list_extract({topic_id+1}, sum_topic_counts_inds, sum_topic_counts_vals)) as n_occurences
FROM doc 
GROUP BY quarter 
ORDER BY quarter""") )

Create a Index from scratch

from sturdystats import Index, Job
import pandas as pd

df = pd.read_parquet('data.parquet')
index = Index(API_key="XXX", name='tech_earnings_calls_2024')

res = index.upload(df.to_dict("records"))
job = index.train(params=dict(), fast=True, wait=True)

Train robust linear models.

pip install sturdy-stats-sdk[regression]

from sturdystats.model import LinearRegressor 
import arviz as az

model = LinearRegression(API_key=API_KEY)
model.sample(X, Y) 
az.plot_trace(model.inference_data)

Detect mislabelled datapoints.

from sturdystats.model import SturdyLogisticRegressor
import arviz as az

model = SturdyLogisticRegressor(API_key=API_KEY)
model.sample(X, Y) 
az.plot_trace(model.inference_data)