CONTACT US

Time series/Longitudinal data analysis

Overview

Survival analysis studies the amount of time it takes before a particular event of interest occurs. It plays a pivotal role in statistical modeling, especially in business, medicine, biology and reliability studies where time-to-event data is fundamental.

What types of questions can be answered with survival analysis?

  • What is the probability of experiencing an adverse outcome from a specific cause by a given time?
  • Does a specific intervention reduce adverse outcomes for all causes or just for specific ones?
  • On average, how long after an intervention/treatment/procedure do different groups experience some specific adverse outcome?

What types of models can we implement at oores Analytics?

  • Kaplan-Meier survival curve with confidence intervals and confidence bands,
  • Wilcoxon, log-rank test,
  • Kernel-Smoothed Hazard Estimator,
  • Cox Proportional Hazards Models,
  • Competing Risks Analysis
  •  

At oores Analytics, we have the tools necessary to analyze and interpret time-to-event data within a rigorous stochastic framework. Talk to us and see how we can help you harness the power of your data, translating same into real-world scenarios, and thus enable you make data-driven decisions!

Case studies

See our case studies about data analyst case

Overview

Time series/longitudinal data analysis involves methods for analyzing data collected over time from the same subjects or units. While both deal with data measured across time, they differ slightly:

  • Time Series Analysis: Focuses on a single subject or system over time (e.g., monthly sales data for a company).
  • Longitudinal Analysis: Involves multiple subjects observed repeatedly over time (e.g., patient health measurements across several visits).

Our time series/longitudinal data expertise includes:

    1. Understanding the data structure
      1. time series: one unit over continuous time (e.g., daily temperature readings)
      2. Longitudinal: repeated measures across individuals (e.g., weight measured monthly in a clinical trial)
    2. Exploratory Data Analysis (EDA)
      1. Visualize trends, seasonality, irregularities
      2. Plot individual and group trajectories
      3. Check for missing values and outliers
    3. Assessing Stationarity (for time series)
    4. Modeling approaches
    5. Address correlation over time
      1. Account for autocorrelation
      2. Use lag structures, random effects, or correlation matrices
    6. Model diagnostics
      1. check residuals for patterns
      2. evaluate goodness-of-fit (AIC/BIC)
      3. cross-validation or forecasting accuracy (e.g., RMSE, MAE)
    7. Prediction & forecasting
      1. Generate short- or long-term forecasts
      2. Include uncertainty bounds
      3. Scenario modeling and intervention analysis (e.g., impact of a policy)

What types of questions can be answered with time series/longitudinal data?

  1. Time series

    • (Trends, seasonal patterns) How do monthly hospital admissions vary over 12 months?
    • (Forecast/predict future values) What will the COVID 19 case count be next month?
    • (Short- and long-term cycles in the data) Are there economic cycles in quarterly health spending?

    Longitudinal

    • (Individual or groups change over time) How does BMI change over 10 years in patients with Type II diabetes?
    • (Time-varying and fixed covariates influence outcomes) Does stress level (measured monthly) predict future depressive symptoms?
    • (Differences in change across subgroups) Do treatment and control groups show different recovery trajectories?
    • (model individual growth or decline) What is the rate of cognitive decline by age and education level?

What time series/longitudinal data models can we implement at oores Analytics?

  1. Time Series Models

    • ARIMA (AutoRegressive Integrated Moving Average)
    • Exponential Smoothing (ETS)
    • Seasonal models (SARIMA)
    • State space models/Kalman filters
    • Prophet (for irregular time series)

    Longitudinal/Panel Data Models

    • Linear Mixed Effect Models (LMM)
    • Generalized Estimating Equations (GEE)
    • Growth curve models
    • Multilevel/hierarchical models, Regression trees
    • Survival analysis (for time-to-event outcomes)

At oores Analytics, we have the tools necessary to analyze and interpret time series/longitudinal data. Talk to us and see how we can help you harness the power of your data, translating same into real-world scenarios, and thus enable you make data-driven decisions!