Categorical Data Analysis 2025/26

Outline of lecture

1. Categorical data - notebook
2. Simpson's paradox - notebook
3. Contingency tables - notebook
4. Discrete distributions - notebook
5. Maximum likelihood estimation - notebook
6. Goodness of fit - notebook
7. Linear regression - notebook
8. Marginal effects - notebook
9. GLM: Count data - notebook
10. GLM: Logistic regression - notebook

Extra materials

• Optimization methods for MLE - notebook
• Interactions and scaling - notebook

Case study

[TBA]

Code files for download

#	Topic	R	Python	Julia	Jupyter
1	Categorical data	.R	.py	.jl	.ipynb
2	Simpson's paradox	.R	--	--	--
3	Contingency tables	.R	.py	.jl	.ipynb
4	Discrete distributions	.R	.py	.jl	.ipynb
5	Maximum likelihood estimation	.R	.py	.jl	.ipynb
6	Optimization methods	--	--	--	.ipynb
7	Goodness of fit	.R	.py	.jl	.ipynb
8	Linear regression	.R	.py	.jl	.ipynb
9	Interactions and scaling	--	--	--	.ipynb
10	Marginal effects	.R	.py	.jl	.ipynb
11	GLM: Count data	.R	.py	.jl	.ipynb
12	GLM: Logistic regression	--	--	--	.ipynb

Problem sets

Submit solutions as a single HTML file via Moodle.

#	Topic	HTML	QMD	Jupyter	Deadline
1	Vacancy analysis (categorical data, distributions, MLE)	html	qmd	ipynb	2026-03-31 23:59
2	Retail store analysis (GOF, linear regression, marginal effects)	html	qmd	--	TBA
3	TBA	--	--	--	TBA

Example final test

Example test -- qmd, html

Required packages / modules

• R:
  • distributions3,
  • maxLik, rootSolve
  • vcd, fitdistrplus
  • marginaleffects, modelsummary
  • car
  • see, performance, patchwork
  • geepack
• Python:
  • scipy, numpy, pandas
  • pingouin, matplotlib, statsmodels
• Julia:
  • Distributions.jl, DataFrames.jl,
  • Optim.jl, Roots.jl
  • HypothesisTests.jl, StatsBase.jl
  • FreqTables.jl, CSV.jl
  • Effects.jl
  • GLM.jl

Description of the data

Source:

• id -- company identifier
• woj -- region (województwo) id (02, 04, ..., 32)
• public -- is the company public (1) or private (0)?
• size -- size of the company (small = up to 9 employees, medium = 10 to 49, big = over 49)
• nace -- NACE (PKD) sections (1 letter)
• nace_division -- NACE (PKD) division (2-digits, https://www.biznes.gov.pl/pl/klasyfikacja-pkd)
• vacancies -- how many vacancies the company reported?

Sample rows from the dataset

           id woj public   size nace nace_division vacancies
    1:  27350  14      1  Large    O            84         2
    2:  26705  14      1  Large    O            84         1
    3: 257456  24      1  Large    O            84         2
    4: 183657  16      1 Medium    O            84         0
    5: 200042  18      1 Medium    O            84         0
   ---                                                      
57476: 244800  08      1 Medium    P            85         0
57477:  62309  08      1 Medium    R            93         0
57478: 106708  08      0 Medium    B            08         0
57479:  62264  08      0 Medium    B            08         0
57480: 255865  08      0  Small    C            23         0

Software versions

R version 4.4.2 (2024-10-31)

Python 3.12.7 | packaged by Anaconda, Inc. | (main, Oct  4 2024, 08:22:19) [Clang 14.0.6 ]

Julia Version 1.11.3
Commit d63adeda50d (2025-01-21 19:42 UTC)