The OBF-Psychiatric dataset is a high-quality collection of wrist actigraphy data from 162 individuals, including patients diagnosed with major depression (bipolar and unipolar), schizophrenia, ADHD, and other mood/anxiety disorders, as well as a healthy control group. It consists of 1565 days of motor activity data with a mean of 9.6 days per individual.

The dataset is ideal for psychiatric research, behavioral analytics, and machine learning tasks such as classification, clustering, and biomarker discovery. It aggregates and standardizes previously published datasets (DEPRESJON, PSYKOSE, HYPERAKTIV) and includes both raw and feature-engineered CSV files.

Actigraphy data was collected using the Actiwatch AW4 device at 32 Hz, then downsampled to 1-minute intervals representing activity intensity. The dataset is homogenized across source studies for easy processing and comparison.

Use cases include:

• Mood state recognition

• ADHD vs schizophrenia discrimination

• Conformal prediction and uncertainty estimation

• Activity-based biomarker research

The dataset is named in honor of Prof. Ole Bernt Fasmer, a pioneer in psychiatric motor activity research.

Introduction

In the high-stakes world of professional football, public opinion often forms around emotions, loyalties, and subjective interpretations. The project at hand aims to transcend these biases by delving into a robust, data-driven analysis of Real Madrid's performance in the UEFA Champions League over the past decade.

Through a blend of traditional statistical methods, machine learning models, game theory, psychology, philosophy, and even military strategies, this investigation presents a multifaceted view of what contributes to a football team's success and how performance can be objectively evaluated.

Exploratory Data Analysis (EDA)

The EDA consists of two layers:

1. Statistical Analysis:

• Set-Up Process: Loading libraries, data frames, determining position relevancy, and calculating average minutes played.
• Kurtosis: Understanding data variance and its internal behavior.
• Feature Engineering: Preprocessing with standard scaler for later ML applications.
• Sample Statistics, Distribution, and Standard Errors: Essential for inference.
• Central Limit Theorem: A focus for understanding by experienced data scientists.
• A/B Testing & ANOVA: Used for null hypothesis testing.

2. Machine Learning Models:

• Ordinary Least Square: To estimate the unknown parameters.
• Linear Regression Models with Sci-Kit Learn: Predicting the dependent variable.
• XGBoost & Cross-Validation: A powerful algorithm for making predictions.
• Conformal Prediction: To create valid prediction regions.
• Radar Maps: For visualizing player performance during their match campaigns.

Objectives

The goal of this analysis is multifaceted: 1. Unveil Hidden Statistics: To reveal the underlying patterns often overlooked in casual discussions. 2. Demonstrate the Impact of Probability: How it shapes matches and seasons. 3. Explore Interdisciplinary Influences: Including Game Theory, Strategy, Cooperation, Psychology, Physiology, Military Training, Luck, Economics, Philosophy, and even Freudian Analysis. 4. Challenge Subjective Bias: By presenting a well-rounded, evidence-based view of football performance.

Conclusion

This project stands as a testament to the profound complexity of football performance and the nuanced insights that can be derived through rigorous scientific analysis. Whether a data scientist recruiter, football fanatic, or curious mind, the findings herein offer a unique perspective that bridges the gap between passion and empiricism.