Survival Analysis

What it does

This skill enables AI agents to perform survival analysis, a statistical method used to analyze the expected duration of time until one or more events happen, such as patient death or machine failure. It specifically handles censored data where the event of interest has not yet occurred for some subjects within the observation period.

When to use it

• Analyzing time-to-event data in clinical trials where patient follow-up periods vary.
• Predicting equipment maintenance schedules based on historical failure logs with incomplete records.
• Evaluating customer churn risks when exit dates are unknown for active accounts.
• Assessing reliability engineering metrics for systems under continuous monitoring.

Key capabilities

• Processing right-censored, left-censored, and interval-censored datasets.
• Calculating Kaplan-Meier survival curves to estimate survival probabilities over time.
• Performing Cox proportional hazards regression to identify factors influencing event occurrence.
• Generating hazard rate plots to visualize risk changes throughout the observation window.

Example prompts

"Analyze this dataset of server crash times, treating servers still running at the end of the month as censored data." "Calculate the median survival time and generate a Kaplan-Meier curve for patients in Group A versus Group B." "Run a Cox regression to determine which variables significantly increase the hazard rate of customer churn in this log."

Tips & gotchas

Ensure your dataset explicitly marks observations as 'event occurred' or 'censored' before analysis begins. Standard mean calculations will yield incorrect results if censored data is not handled specifically by survival functions.