CellMigrationLab / Plot-Stats

Jupiter notebooks created to help us plot and analyse our datasets

Plot&Stats

Jupyter notebooks created to help us plot and analyse our datasets

Quick Start

Access the notebooks directly in Google Colab for an easy-to-use environment:

• Plot&Stats - Wide to Tidy Format: Transform wide-format data into tidy format for analysis.

  

• Plot&Stats - BoxPlots: Enhanced data visualization, quantifies effect size, adapts to non-standard distributions, streamlines analysis, ensures equitable group representation, achieves dataset balance for fairer comparisons, and delivers in-depth insights from balanced data.

  

• Plot&Stats - dimensionality reduction: Notebook for generating PCA, UMAP or t-SNE dimensional reduction of multidimensional datasets.

  

About the Notebooks

Plot and Stats - wide to tidy

This notebook is designed to transform wide-format data into a tidy format for further analysis

Wide and tidy formats represent two principal ways of structuring tabular data:

• Wide Format:

  • Each row represents a subject or item.
  • Observations spread across multiple columns.
  • Suitable for data entry or presentation.
  • Example with biological repeats:

    | Subject | Cond1_Repeat1 | Cond1_Repeat2 | Cond2_Repeat1 | Cond2_Repeat2 |
    |---------|---------------|---------------|---------------|---------------|
    | 1       | ValueA        | ValueB        | ValueC        | ValueD        |
    

• Tidy Format:

  • Each column is a variable, each row an observation.
  • Suited for statistical analysis and plotting.
  • Each row represents a unique combination of variables.
  • Example with biological repeats:

    | Subject | Condition | Repeat | Value  |
    |---------|-----------|--------|--------|
    | 1       | Cond1     | 1      | ValueA |
    | 1       | Cond1     | 2      | ValueB |
    | 1       | Cond2     | 1      | ValueC |
    | 1       | Cond2     | 2      | ValueD |
    

Wide format is more readable for direct comparisons across a subject's measurements, while tidy format is optimized for analysis, making data transformations, summarizations, and visualizations more straightforward.

Plot&Stats - BoxPlots

This Jupyter Notebook is crafted for the purpose of analyzing datasets maintained in a tidy format. It integrates a comprehensive set of functionalities for in-depth data examination, statistical evaluation, and dataset balancing, enhancing both the analysis and interpretability of your data.

Key Features

• Boxplots with Labels: Creates detailed boxplots that visually differentiate each data point and clearly label repeats, facilitating an immediate understanding of the data distributions.

• Cohen's d Calculation: Enables the computation of Cohen's d value, offering a quantitative measure of the effect size between groups and highlighting the significance of observed differences.

• Randomization Test Based on Cohen's d: Implements a non-parametric randomization test using Cohen's d, suitable for datasets that may not meet the strict assumptions required for traditional parametric tests. More info on randomization tests here.

• Statistical Summaries Export: Automatically generates and exports comprehensive statistical summaries, providing a snapshot of crucial metrics throughout the dataset.

• Dataset Balance Check: Examines the dataset for balance across various conditions and repeats, ensuring that each group is equally represented in subsequent analyses.

• Dataset Resampling: Facilitates the adjustment of the dataset to a balanced condition through downsampling, making comparisons across groups fairer and more meaningful.

• Analysis of Resampled Dataset: Offers tools to further analyze the balanced dataset, with plots and statistical tests designed to uncover robust insights from the equitably represented data.

This notebook acts as a powerful tool for researchers and data analysts, streamlining the workflow from data ingestion to comprehensive analysis, thus enabling a deeper and more accurate exploration of datasets.

Plot&Stats - dimensionality reduction

Key Features

• PCA Analysis & Plots: Generates PCA plots that visually represent the data's variance along principal components, along with the PCA loadings to identify contributing features.
• UMAP or t-SNE Visualization: Utilizes UMAP or t-SNE for dimensionality reduction to project high-dimensional data into a lower-dimensional space, enhancing cluster identification.
• HDBSCAN Clustering: Applies the HDBSCAN algorithm to identify naturally occurring clusters in the data without specifying the number of clusters a priori.
• Fingerprinting Plots: Creates fingerprinting plots that detail the distribution of the identified clusters accross the conditions.
• Boxplots of Clusters: Generates boxplots for each identified cluster to compare distributions across different conditions.