Skip to content

Milestone 5 - Modeling

🎯 Objectives

By the end of this milestone, you should be able to:

  • Specify and estimate a linear regression model using statsmodels.
  • Interpret regression coefficients in relation to your hypothesis.
  • Export regression tables and visualisations into a research paper.

🧠 Tips for Success

  • Use the formula API. Specify models using DV ~ IV + CV_1 + CV_2 syntax.
  • Interpret with context. Focus on what the coefficients mean for your hypothesis, not just statistical significance.
  • Compare across models. Notice how coefficients change as you add controls.

🧩 The Milestone

Building on Milestone 4's exploratory analysis, you will now test your hypotheses using simple regression models. This milestone emphasizes moving from visualization to statistical inference.

You need to:

  • Use your previous Milestone 4 (.typst file and Python script) as a starting point, incorporating any feedback received.

  • Apply linear regression techniques learned in:

  • Estimate multiple regression models side by side with progressively added control variables (covariates) to test your hypothesis and observe coefficient stability.
  • Export your regression table using the make_table() function from mmisc.
  • Interpret your results in relation to your theoretical expectations.

What is expected from a manuscript perspective

Your Milestone 5 typst file is an updated Milestone 4 typst file that:

  • Does not exceed 4000 words (excluding references/bibliography).
  • Integrates a regression table: Use make_table() to create your regression summary table.
  • Includes a results section:
    • Assesses coefficient stability: Note how IV coefficients and the R-squared values changes across models.
    • Connects results to hypothesis: Explain what the coefficients reveal about your theoretical expectations (SSS).
    • Discusses limitations: Acknowledge potential model limitation and potential threats to causal inference (e.g., omitted variable bias, reciprocal causation, ...)
    • Revisit your DAG (optional): If you created a DAG for Milestone 2, you can use it to help you reflect on how your thinking has evolved. No need to create a new DAG unless you want to.

What is expected from a coding perspective

Similarly, your Milestone 5 Python script is an updated Milestone 4 script that contains:

  1. Model Specification

    • Baseline model: Estimates a simple regression with your main IV predicting the DV
    • Control model: Estimates a model with theoretically relevant control variables (CVs) predicting the DV but WITHOUT your main IV
    • Full model: Estimates a full model with CVs and your main IV predicting the DV

  2. Model Comparison

    • List of models: Create a list of models that you will use with make_table() and summary_col()
    • Use summary_col(): Compare coefficients and estimates across models in your interactive Python terminal.
    • Export models with make_table(): Generate a publication-ready table for your typst document

  3. Visualization (Bonus)

    • Use a modern visualization library (such as Altair)
    • Create a plot to visualize the effects of your final model result (e.g. coefficient plot with confidence intervals).
    • Discuss/comment the figure and integrate it into your paper.

💾 Submission Guidelines

  1. Export final files from both Typst and your Python script:
    • da-milestone5-group0.typ
    • da-milestone5-group0.pdf
    • da-m5-script-group0.py
  2. Keep file names consistent for organization (e.g., da-milestone5-group1.typ, da-milestone5-group1.pdf, da-m5-script-group1.py).
  3. Email all 3 files attached to the instructor before the deadline with the subject line: 
    [DATA-ANALYSIS] Milestone 5 - Group X
    (List all contributing group members in the body of the email).
  4. Manuscript length must not exceed 4000 words (excluding references).

This milestone marks the transition from exploratory analysis to hypothesis testing, bringing you closer to your final research paper. Well, your final research paper is just going to be some minor finishing touches.