VIBE CODE COMPARISON - Best agent for data analysis (VS Code)

WHAT IS THIS VIDEO ABOUT? In this video I show you how to use an agent in Visual Studio Code to populate a Jupyter notebook with analysis of a data file. The video shows you how to go about it and then compares the results of using Gemini 2.5 Pro, Claude 4 Sonnet, GPT-4.1, and o4-mini as LLM. WHY IS THIS USEFUL? Not all coding models are created equal. One particular model stands out and creates to best Jupyter notebooks. WHAT WILL YOU NEED? The data file named Metrics.csv is available at https://github.com/juanklopper/TutorialData You need to have a version of Python installed on your system. You can do so from https://python.org You also need Visual Studio Code installed from https://code.visualstudio.com/download The Python, Jupyter, and Github Copilot extensions must be installed in . THE AGENT INSTRUCTION FILE Save the file to .github/instructions and name it notebook.instructions.md --- applyTo: "notebook" --- # Notebook Instructions * Start all notebooks with the Python code `%config InlineBackend.figure_format = 'retina'` to enable high DPI plotting. * Use a clear and descriptive title for the notebook. * Use markdown section cells to organize the notebook into sections. * Import all necessary libraries at the beginning of the notebook. Install any Python packages that are required and not yet installed * Use code cells for executable code. * Ensure that the code is well-commented and easy to understand. * Do not create excessively long code cells. * Break code cells into shorter, manageable chunks. * Do not use `print` statements to comment on the results of code. Use text and LaTeX in markdown cells after each code cell execution instead. * Use markdown cells with text and LaTeX for results, interpretations, explanations, comments, and documentation. For instance if the code cell contained the code `df.Data.mean()` and the result is 42, use a markdown cell to write: The sample mean $\bar{X}=42$ beats per minute. * Use consistent formatting throughout the notebook for a professional appearance. THE PROMPT Remember to select Agent and not Ask. The CSV file contains data on an experiment to determine the sensitivity, specificity, and the positive and negative predictive values of a new test for a specific disease. A total of 322 participants were selected for the experiment. To reflect the current prevalence of the disease in the community, which is 18.6%, 60 participants have the disease and 262 do not. The `Disease` column has two classes. `No` indicates the absence of the disease as measured by a gold-standard test, and `Yes` indicates the presence of the disease by the same gold-standard test. The `Test` column has two classes and records the results of the new test. `Negative` indicates a negative test for the disease and `Positive` indicates a positive rest for the disease. Analyze the data using the following plan: 1. Calculate and visualize the frequency and relative frequency of each class in each of the `Disease` and the `Test` columns. 2. Create a contingency table of the two columns and visualize the results. 3. Calculate and interpret the joint probabilities of the contingency table. 4. Calculate and interpret the marginal probabilities of contingency table. 5. Calculate and interpret the sensitivity and the specificity of the new test. 6. Calculate and interpret the positive and the negative predictive values of the new test. 7. Write a summary and conclusion of the analysis of the new test.