Clinical Biostatistics Modeling

Survival Analysis • Cox Models • Generalized Additive Models (GAM)

This repository contains modular R scripts for core biostatistical workflows used in clinical and biomedical research.
The code demonstrates reproducible methods for time-to-event analysis, hazard modeling, and nonlinear effect estimation using Generalized Additive Models (GAM).

Important:
These scripts were originally developed as part of projects using sensitive biomedical datasets (clinical imaging, psychiatric cohort data, or patient-level outcomes).
Due to GDPR and data-sharing restrictions, the original datasets cannot be made public.

The versions provided here include generic, synthetic, or user-replaceable data-loading placeholders, ensuring the methods are reproducible without exposing confidential data.

---

📦 Repository Contents

1. survival_analysis_v2.R

A complete time-to-event modeling pipeline using R.

Includes:

• Kaplan–Meier estimator
• Log-rank test
• Cox proportional hazards model (univariate + multivariate)
• Hazard ratio extraction and visualization
• Time-to-event prediction workflow
• Clean separation of loading, modeling, and plotting functions

Key Packages:
survival, survminer, dplyr, ggplot2

---

2. code_GAM_v2.r

Script for fitting Generalized Additive Models (GAM) to clinical/biomedical variables.

Includes:

• Model specification with nonlinear smooth terms
• Covariate adjustment
• Partial effect plots (smooth splines)
• Handling continuous and categorical predictors
• Clear model summary + interpretation outputs

Key Packages:
mgcv, ggplot2, dplyr

---

🧠 Why These Methods Matter in Biomedical Research

Survival Analysis

Used for:

• Clinical progression
• Hospitalization risk
• Transition prediction (e.g., CHR → psychosis)
• Treatment response duration
• Risk factor modeling

Provides:

• Hazard ratios
• Time-dependent probabilities
• Survival curves
• Clinically interpretable risk profiles

---

Generalized Additive Models (GAM)

Used for:

• Nonlinear biomarker–outcome relationships
• Brain/clinical covariate effects
• Age effects with nonlinearity
• Modeling without unrealistic linear assumptions

GAMs are especially helpful in:

• Neuroimaging
• Clinical psychiatry
• Biomarker discovery
• Developmental and lifespan modeling

---

🔐 Data Privacy & Compliance Notice

This repository includes code only, not data.
The original analyses were performed on GDPR-protected biomedical datasets, including patient-level clinical, imaging, or questionnaire data.

To comply with:

• GDPR (EU)
• Institutional Ethics approvals
• Data sharing agreements

➡️ No raw data or identifiable variables can be shared.

All current examples are built to work with:

• synthetic data
• user-provided datasets
• de-identified tables

You may adapt the scripts to your own dataset structure.

---

▶️ How to Use

1. Install dependencies

install.packages(c("survival", "survminer", "mgcv", "ggplot2", "dplyr"))
---