Algorithms, Evidence and Data Science

Course Leader

Prof. Dr. Frank Pijpers

Target Group

Staff of national statistical institutes involved in the production process who want to acquire a good understanding of modern computational statistics and algorithmic approaches to data analysis

Entry Qualifications

• Sound command of English. Participants should be able to make short interventions and to actively participate in discussions

• Familiarity with statistical software (R, Python, or Excel) is beneficial.

Objective(s)

• Explain the principles behind key algorithms in data science and statistical modelling.

• Apply basic computational techniques such as resampling, Monte Carlo, and bootstrap for inference.

• Evaluate strengths, limitations, and assumptions of statistical algorithms in the context of official statistics.

• Understand how computation has transformed statistical practice.

• Recognize the role of reproducibility, ethics, and transparency in algorithmic data science.

Contents

1. Classical statistical inference

   1. Frequentist, Bayesian

   2. Maximum Likelihood Estimation, parametric models

   3. model selection

2. Basic data science methods

   1. ridge regression 

   2. regression trees

   3. jackknife & bootstrap

   4. Markov Chain Monte Carlo (MCMC) methods

3. Supervised & unsupervised learning

   1. logistic regression, random forests, neural networks. 

   2. clustering, dimensionality reduction

   3. neural networks & deep learning

   4. kernel methods

4. Future directions

   1. algorithmic output as evidence: uncertainty, variability, and bias.

   2. AI and machine learning integration.

   3. online learning and e-values

   4. Reproducibility, transparency and ethical considerations.

Expected outcome

Once the course is completed, the participants would be expected to have a more thorough understanding of modern (computational) statistical methodology, an ability to read the relevant recent scientific and methodological literature and apply it to their work. This would equip NSI staff to interpret algorithmic outputs responsibly, make evidence-based decisions, and understand the evolving landscape of statistical methodology.

Training Methods

• Presentations and lectures, practical examples

• Exchange of views/experiences on national practices

• Exercises

Required Reading 

n/a

Suggested Reading

B.Efron and T.Hastie: Computer Age Statistical Inference: Algorithms, Evidence and Data Science, CU Press, 2016

Possibly: Richard McElreath: Statistical Rethinking, CRC Press, 2020

Required Preparation

Participants should supply a brief document with information about their education, position, organisation and expectations of the course (max 200 words)

Trainer(s)/
Lecturer(s)

Prof. Dr. Frank P. Pijpers (Statistics Netherlands)

Dr. Rosanne Turner (Statistics Netherlands) – Guest speaker