Welcome to the homepage of Frank Miller

I am Professor in Statistics at the University of Stockholm where I work since 2013. Prior to my current position, I applied and developed statistical methods in the pharmaceutical industry for 10 years (at AstraZeneca in Södertälje, Sweden). I received my Ph.D. in 2002 from the University of Karlsruhe (Germany).
In my statistical research I focus on innovative methods needed in practice and feasible to implement. My research interests include:
Achievement tests, active machine learning, adaptive and sequential designs, biostatistics, clinical trials, optimal experimental designs, optimization algorithms.

Recent and upcoming teaching at Stockholm University, Department of Statistics

  • Jan/Feb 2022: Computational statistics
  • Jan 2022: One lecture about Computer based methods: randomization, bootstrap within the course Biological statistics III
  • Nov 2021: Experimental design
  • May 2021: Two lectures about Active machine learning within the course Machine learning
  • Mar/Apr 2021: Ph.D. course Optimisation algorithms in statistics II
  • Jan/Feb 2021: Computational statistics
  • Jan 2021: One lecture about Computer based methods: randomization, bootstrap within the course Biological statistics III
  • Oct/Nov 2020: Ph.D. course Optimisation algorithms in statistics I
  • Sept 2020: Experimental design
More about my teaching

Supervision of PhD projects

Currently, I am supervising the PhD student Karl Sigfrid as main supervisor. His research topic is: Design of computerized adaptive tests.

Topics for future PhD projects
  • Improved methods for pretesting achievement tests and implementation. Questions for larger achievement tests like PISA, högskoleprovet, and national tests in school need to be pretested in advance. The Swedish research council funded a reseach project to improve methods for this pretesting. There is possibility for a further PhD student to contribute to this project.
  • Optimization algorithms. In several areas of statistics including optimal experimental design and machine learning, it is essential to have efficient algorithms for computing optimal solutions numerically. While optimization algorithms have a considerable history, many new algorithms have been suggested in recent years. In this PhD project, we will work on improving modern optimization algorithms.
  • Active machine learning. This area deals with situations where unlabeled data is available but there is the possibility to label some of the observations. Methods of optimal experimental design give the opportunity to choose observations for labeling which are most suitable. The PhD project aims to improve the current active machine learning methods.

Previous PhD projects

I was co-supervisor for Oskar Gustafsson from 2015-2020. Pär Stockhammar was main supervisor; here the PhD thesis.

Participation in research projects
  • 2020 to 2023: Optimal calibration of questions in computerized achievement tests. Grant by the Swedish Research Council (Vetenskapsrådet). I am main applicant. Link to the project homepage. Here an article in Swedish on the Department's homepage from the start of the project.
  • 2014 to May 2017: Innovative methodology for small populations research (InSPiRe). Grant by European Union's Seventh Framework Programme. Main applicant: Nigel Stallard (University of Warwick). I was co-applicant. Link to the project homepage.
  • 2015 and 2016: Integrated DEsign and AnaLysis of small population group trials (IDEAL). Grant by European Union's Seventh Framework Programme. Main applicant: Ralf-Dieter Hilgers (University Hospital Aachen). I contributing to a workstream lead by co-applicant Carl-Fredrik Burman (AstraZeneca). Link to the project homepage.

Selected publications

More about my research with a complete list of my publications

Recent conference and seminar talks

  • Optimal dose finding for efficacy-safety-models of Emax-type
    Conference: DAGStat (Hamburg, Germany, 2022)
  • Optimal pretesting of questions in large achievement tests
    Seminar talk at KU Eichstätt-Ingolstadt, Department of Mathematics (Eichstätt, Germany, 2019)
  • Optimal item calibration designs for computerized achievment tests
    Conference: DAGStat (Munich, Germany, 2019)
  • Applied decision theoretic designs for clinical studies
    Seminar talk at Stockholm University, Department of Statistics (Stockholm, Sweden, 2018)
  • Optimal design for dose-finding in clinical trials
    Seminar talk at Linköping University, Division of Statistics and Machine Learning (Linköping, Sweden, 2017)
  • Applied decision theory for clinical trials in small populations
    Conference: Nordstat (Copenhagen, Danmark, 2016)
  • Optimizing the design of a dose-finding trial: theoretical methods and practical aspects
    Conference: EFSPI European Statistical Meeting on Dose Selection in Late Stage Clinical Development (Brussels, Belgium, 2015)
  • Adaptive dose-finding with control of familywise error rate and power
    Conference: Multiple comparisons procedures (Hyderabad, India, 2015)
Complete list of my talks at conferences or research seminars