A practical exploration of statistical methods using Python.

---

This project explores a synthetic dataset of 2,000 observations and 22 variables using Python. It covers the main stages of a statistical analysis workflow, from data cleaning and descriptive statistics to hypothesis testing, correlation, and regression.

The analysis is organized into reusable Python modules, with a Jupyter notebook bringing the methods and results together. The dataset includes missing values, outliers, and several types of variables, making it useful for exploring how different statistical methods are applied and interpreted.

Dataset note

All data in this project is synthetic. The findings illustrate statistical methods on simulated data and should not be interpreted as evidence about real populations or medical outcomes.

01 · Project at a glance

2,000 OBSERVATIONS

22 VARIABLES

Python CORE LANGUAGE

Modular CODE + NOTEBOOK

02 · Selected findings

Research question	Method	Result
Do post-treatment scores differ between the drug and placebo groups?	Welch's independent t-test	The drug group had higher scores in this dataset (t = 12.37, p < .001, Cohen's d = 0.55).
Do scores differ before and after treatment?	Paired t-test	Post-treatment scores were higher on average (t = −34.62, p < .001, d = 0.77).
Does performance vary across groups A, B, and C?	One-way ANOVA and Tukey HSD	The overall group difference was significant (F = 105.52, p < .001, η² = 0.096); all three pairwise comparisons were significant.
Are smoking status and disease status associated?	Chi-square test of independence	The variables were significantly associated (χ² = 180.34, p < .001).
How is weekly study time related to exam scores?	Simple linear regression	Each additional weekly study hour was associated with an estimated 2.81-point increase in exam score (R² = 0.671).
Which variables predict passing the exam?	Logistic regression	Exam score was a significant predictor (odds ratio = 1.08 per point), while weekly study hours were not significant after exam score was included in the model.

These results describe the simulated dataset, not causal effects in real-world settings. Statistical significance is reported alongside effect sizes where available, since a small p-value alone does not show the magnitude or practical importance of a difference.

The notebook also compares selected parametric tests with non-parametric alternatives. The results agree for the group comparisons examined, providing a useful cross-check rather than a guarantee that every model assumption is satisfied.

03 · Visual results

Correlation structure Relationships between numerical variables

Group comparison Distribution of outcomes across groups

04 · Dataset

The dataset contains 2,000 rows and 22 columns, combining numerical and categorical variables.

Variable type	Examples
Nominal	gender, education, city, treatment, group
Continuous	age, height_cm, weight_kg, income, exam_score
Ordinal	job_satisfaction_1to5
Count	children_count
Binary	passed_exam, smoker, has_disease
Date	register_date

Missing values are included in selected columns, and outliers are present in variables such as income. The analysis examines these issues before proceeding to statistical testing. The main notebook loads data/comprehensive_dataset.csv.

05 · Methods

Data preparation and descriptive statistics

• Missing-value reporting and imputation
• IQR-based outlier detection and outlier capping
• Mean, median, mode, variance, skewness, and kurtosis
• Frequency tables and confidence intervals
• Shapiro–Wilk normality tests, with attention to their sensitivity at large sample sizes

Hypothesis testing and effect sizes

• One-sample, independent-samples, and paired t-tests
• One-way ANOVA and Tukey HSD post-hoc comparisons
• Chi-square tests of independence
• Mann–Whitney U, Wilcoxon signed-rank, and Kruskal–Wallis tests
• Cohen's d and eta-squared

Correlation, regression, and visualization

• Pearson and Spearman correlation coefficients
• Simple and multiple ordinary least squares (OLS)
• Regression diagnostics and logistic regression with odds ratios
• Distribution plots, group comparisons, scatter plots, and correlation heatmaps

06 · Repository map

View project files

statistical-analysis-project/
├── data/
│   ├── cleaned_dataset.csv
│   ├── comprehensive_dataset.csv
│   ├── raw_dataset.csv
│   └── generate_raw_dataset.py
├── notebooks/
│   └── statistical_analysis.ipynb
├── outputs/
│   └── figures/
├── src/
│   ├── __init__.py
│   ├── correlation_regression.py
│   ├── data_cleaning.py
│   ├── data_loader.py
│   ├── descriptive_stats.py
│   ├── hypothesis_tests.py
│   ├── regression.py
│   └── visualization.py
├── tests/
│   └── test_statistics.py
├── requirements.txt
├── LICENSE
└── README.md

07 · Getting started

1. Clone the repository and install dependencies

git clone https://github.com/ParnazAli/statistical-analysis-project.git
cd statistical-analysis-project
pip install -r requirements.txt

2. Launch the notebook

Run these commands from the repository root:

cd notebooks
jupyter notebook statistical_analysis.ipynb

3. Run the tests

Open a separate terminal at the repository root:

pytest tests/ -v

08 · Tech stack

---

Parnaz Ali

Email · Repository

_{Released under the MIT License · See LICENSE for details.}