Tutorial for Performance Table (gt)

[uriahf]

A tutorial for different options for performance table and building the table from scratch.

Use cases:

• One Population, One Model
• One Population, Several Model
• Several Populations, one model for each

Strata By:

• Probability Threshold

• Predicted Positives

• Expendable Confusion Matrix

• Column groups for each population / model

[uriahf]

Goals:

For me:

• a good opportunity to document the logic behind the different use cases
• a nice teaser for {rtichoke}

For the user:

• Embedding html
• Embedding latex
• Embedding expendable
• Using colors
• Using background colors
• Using Markdown
• Best practices for tables
• Using fonts (google fonts?)

[uriahf]

1. Prepare Performance Data 🧮
2. Use gt package 📦
3. Add colors to title and subtitle 🎨🖌️
4. Expandable rows for confusion matrix ↕️
5. Add latex (?)✒️
6. Compare Several Models 🥇🥈🥉 🌲🙈🔀
7. Several Populations 🙋🙋‍♂️
8. One line for each row ↔️

[uriahf]

• Confusion Matrix in css details tag: How to make a note for column group? How to sort borders for rows

• Equations for performance metrics: latex? Html with classes?

[uriahf]

• Where are the NPV / PPV values? (Replace nan with "")

[uriahf]

Example:

Based on https://emilhvitfeldt.github.io/ISLR-tidymodels-labs/tree-based-methods.html

Logistic Regression - regular syntax

[uriahf]

Set code download to TRUE

[uriahf]

Every theoretical concept: no more than four sentences, always link to other sources:

• Fairness (Several Populations)
• Overfit and Underfit (stat quest?) (Several Populations)
• Backwards time-flow information - prediction and not detection.

[uriahf]

• Theoretical Concepts
• {gt} functions