3-D bridge now opened!

The 3D printed bridge I had mentioned in a previous blog was installed and officially opened over an Amsterdam canal, last month. Besides the company printing this steel bridge, over six months, researchers from the Turing Institute, including my friend Mark Girolami, developed a virtual version of the bridge using embedded sensors to monitor the behaviour and evolution of the real bridge over time and usage. Next time I am in the city, I’ll make sure to run bridge sampling!

Adrian Raftery’s course in Paris

UW Professor and U.S. National Academy of Sciences member Adrian Raftery has received the 2020 FSMP research chair and as a result will be visiting Paris this Fall 2021. He will be located at the MAP5 laboratory at the University of Paris. In particular, he will give a 20-hour Master course on statistical semography. This will be given over four successive Tuesdays, with 5 hours of lectures per week. The course is open to all. Attendance is free of charge but registration is mandatory. (To register, please fill the attached form. Lectures will be given in the salle du conseil, on the 7ft floor of the Saint-Germain-des-Prés campus, 45 rue des Saints-Pères, 75006 Paris.)

Demography aims to estimate and forecast population, fertility, mortality and migration. This is important for government policy-making, private sector planning, and research in the health and social sciences, and also critical for climate science and global health. It has traditionally been done using deterministic mathematical methods, but these ignore uncertainty and measurement error.  In the past decade, modern statistical methods were developed for this by our group at the University of Washington, and these were recently adopted by the  United Nations for their official population forecasts for all countries.  Statistical demography is expanding rapidly,  and this course will teach theory and practice of  methods and models of the field, with a focus on current and potential future research.

The topics will be:

1. Review of basic mathematical demographic methods.

2. Modeling age-specific rates, including model schedules and Lee-Carter method.

3. Statistical modeling and projection of fertility, mortality, migration and population.

4. Reconstructing population and vital rates from imperfect data.

likelihood inference with no MLE

“In a regular full discrete exponential family, the MLE for the canonical parameter does not exist when the observed value of the canonical statistic lies on the boundary of its convex support.”

Daniel Eck and Charlie Geyer just published an interesting and intriguing paper on running efficient inference for discrete exponential families when the MLE does not exist.  As for instance in the case of a complete separation between 0’s and 1’s in a logistic regression model. Or more generally, when the estimated Fisher information matrix is singular. Not mentioning the Bayesian version, which remains a form of likelihood inference. The construction is based on a MLE that exists on an extended model, a notion which I had not heard previously. This model is defined as a limit of likelihood values

called the MLE distribution. Which remains a mystery to me, to some extent. Especially when this distribution is completely degenerate. Examples provided within the paper alas do not help, as they mostly serve as illustration for the associated rcdd R package. Intriguing, indeed!

 

MCqMC 2022 im Linz

The next MCqMC conference is still scheduled to take place in beautiful Linz, Austria, on 17-22 July, hopefully for real rather than virtually. (The above picture was taken in 2014, while running to the top of a nearby hill, the Pöstlingberg, featuring this Wallfahrtsbasilika. ) Details are already available on the conference webpage.

content which deviates from the norm [from Pest county]