Statistics seminar

Statistics seminars are informal events where new ideas can be presented and discussed. The aim is to provide a venue for discussion and exchange of ideas between different statistics research groups and others interested in statistics. All seminars are in English. The program is updated continuously so look out for new titles.

Note! The Spring term seminars are cancelled for the time being.

Time: Fortnightly on Tuesdays at 12:15 - 13:00.

Place: Exactum C124.

All interested are warmly welcome!

  • 28.1. Tuomo Nieminen
    Vaccine safety surveillance using national health registers.
  • 11.2. Leo Lahti
    Statistics, ecology, and the universal neutral theory of biodiversity
  • 25.2. Ulrike Baum
    Mitigation of biases in register-based cohort studies
  • 3.3. Kari Auranen
    Sunset for statistical significance
  • 17.3. Leena Kalliovirta
    Time series modeling of economic data
  • 31.3. Aapeli Nevala
  • 14.4. Mikko Heikkilä
    Differentially private MCMC


  • 28.4. Jorma Kilpi
    Data-analysis of vehicular traffic.
  • 12.5.
  • 19.2. Juho Kettunen
    Joint modelling of the vegetation cover using Gaussian processes
  • 5.3. Sangita Kulathinal
    Time-to-event study designs and analysis
  • 21.3. Mark Girolami (Note unusual time and place!)
    Probabilistic Numerical Computation: a Role for Statistical Science in Numerical Analysis?

    Abstract: Consider the consequences of an alternative history. What if Leonhard Euler had happened to read the posthumous publication of the paper by Thomas Bayes on “An Essay towards solving a Problem in the Doctrine of Chances”? This paper was published in 1763 in the Philosophical Transactions of the Royal Society, so if Euler had read this article, we can wonder whether the section in his three volume book Institutionum calculi integralis, published in 1768, on numerical solution of differential equations might have been quite different.

    Would the awareness by Euler of the “Bayesian” proposition of characterising uncertainty due to unknown quantities using the probability calculus have changed the development of numerical methods and their analysis to one that is more inherently statistical?

    Fast forward the clock two centuries to the late 1960s in America, when the mathematician F.M. Larkin published a series of papers on the definition of Gaussian Measures in infinite dimensional Hilbert spaces, culminating in the 1972 work on “Gaussian Measure on Hilbert Space and Applications in Numerical Analysis”. In that work the formal definition of the mathematical tools required to consider average case errors in Hilbert spaces for numerical analysis were laid down and methods such as Bayesian Quadrature or Bayesian Monte Carlo were developed in full, long before their independent reinvention in the 1990s and 2000s brought them to a wider audience.

    Now in 2017 the question of viewing numerical analysis as a problem of Statistical Inference in many ways seems natural and is being demanded by applied mathematicians, engineers and physicists who need to carefully and fully account for all sources of uncertainty in mathematical modelling and numerical simulation.

    Now we have a research frontier that has emerged in scientific computation founded on the principle that error in numerical methods, which for example solves differential equations, entails uncertainty that ought to be subjected to statistical analysis. This viewpoint raises exciting challenges for contemporary statistical and numerical analysis, including the design of statistical methods that enable the coherent propagation of probability measures through a computational and inferential pipeline.

  • 2.4.  Ulpu Remes
  • 16.4. Jalo Nousiainen
  • 30.4. Anna Suomenrinne-Nordvik
  • 14.5. Umberto Simola
    Machine Learning Accelerated Likelihood-Free Event Reconstruction in Dark Matter Direct Detection
  • 25.9. canceled
  • 9.10. Joseph Sakaya
    Lambert Matrix Factorization
  • 23.10. Hande Topa
    Gaussian process modelling of genome-wide high-throughput sequencing time series
  • 6.11. Tomasz Kusmierczyk
    Loss-calibrated variational inference
  • 20.11. Gleb Tikhonov
    Computationally efficient joint species distribution modeling of big spatial data – advances and challenges
  • 4.12. Malcolm Itter
  • 18.12. Elina Numminen
    The fall of the Finnish flying squirrel
  • 23.1. Mikko Heikkilä
    Differentially private Bayesian learning on distributed data
  • 6.2. Umberto Simola
    An Adaptive Approximate Bayesian Computation Tolerance Selection Algorithm
  • 20.2. School holidays break
  • 6.3. Jussi Mäkinen
    Hierarchical Bayesian framework for inferring species densities from heterogeneous observations
  • 20.3. Liisa Iivonen
    Novel Bayesian models for past climate reconstruction from pollen records
  • 3.4.
  • 17.4. Jia Liu
    Bayesian model based spatio-temporal sampling design with an application on species distribution modeling
  • 15.5. Timothy E. O’Brien at 10-12 (Notice the unusual time)
    Statistical Modelling and Design: From Theory to Practice

    Researchers often find that nonlinear regression models are more applicable for modelling various biological, physical and chemical processes than are linear ones since they tend to fit the data well and since these models (and model parameters) are more scientifically meaningful. These researchers are thus often in a position of requiring optimal or near-optimal designs for a given nonlinear model. A common shortcoming of most optimal designs for nonlinear models used in  practical settings, however, is that these designs typically focus only on (first-order) parameter  variance or predicted variance, and thus ignore the inherent nonlinear of the assumed model  function. Another shortcoming of optimal designs is that they often have only support points, where  is the number of model parameters.
    Furthermore, measures of marginal curvature, first introduced  in Clarke (1987) and extended in Haines et al (2004), provide a useful means of assessing this  nonlinearity. Other relevant developments are the second-order volume design criterion introduced in Hamilton and Watts (1985) and extended in O’Brien (2010), and the second-order MSE criterion developed and illustrated in Clarke and Haines (1995).
    In the context of applied statistical modelling, this talk examines various robust design criteria and those based on second-order (curvature) considerations. These techniques, coded in popular software packages, are illustrated with several examples including one from a preclinical dose-response setting encountered in a recent
    consulting session.

  • 19.9. Jarno Vanhatalo
    Spatial Clustering Using Gaussian Processes Embedded in a Mixture Model
  • 3.10. Joonas Jälkö
    Privacy aware variational inference
  • 17.10. Marcelo Hartmann
    Approximate inference for location-scale Gaussian process regression with Student-t probabilistic model
  • 24.10. Juha Karvanen (Notice! this is extra between two normal schedule seminars)
    Towards automated causal inference
  • 31.10. Johan Pensar
    Structure learning of context-specific graphical models
  • 14.11. no seminar
  • 28.11. Tommi Mäklin
    Probabilistic quantification of bacterial strain mixtures
  • 12.12. Christian Benner
    Efficient variable selection among thousands of correlated genetic variants using summary data from genome-wide association studies