University of Minnesota, Twin Cities
School of Statistics
Charlie Geyer's Personal Home Page
Also I must abide by the
University Accessibility of Information Technology Policy, which I try to do, as well as I can. I'm all for
accessibility, but it is hard to test.
My worst accessibility offense is using plots in statistics (including over
the web via Rweb). I don't know
how to make them accessible. There's a good research area for anyone
interested in statistical graphics. Apparently, this is already an
area of some
research interest, but we are a long way from an R device driver
that outputs accessible statistical graphics (AFAIK).
- Seminars, preprints, tech reports, conference proceedings.
- On Saturday, November 5, 2022 I gave a talk about reproducibility
at a workshop Mapping the Frontiers of Reproducibility: Foundational
Questions held by the Minnesota Center for Philosophy of Science.
- On Thursday, September 8, 2022 I gave a departmental seminar about
how to write and talk about mathematics.
Here are the slides.
- Submitted paper Computationally efficient likelihood inference in exponential families when the maximum likelihood estimator does not exist from Dan Eck's thesis.
- On Tuesday, June 19, 2018 I gave a talk at the Mini-Conference
to Celebrate Elizabeth Thompson's Contributions to Statistics,
Genetics and the University of Washington. The title of
the talk was Fast Valid Statistical
Inference when the Maximum Likelihood Estimate Does Not Exist in an
Exponential Family Model and the
Usual Asymptotics are Bogus.
The page linked above also links to supplementary material showing how the
figures in the slides were produced. Both slides and supplementary
material are reproducible (knitr source provided).
- Variable Transformation to Obtain
Geometric Ergodicity in the Random-walk Metropolis Algorithm
(with Leif Johnson)
paper submitted to Annals
of Statistics and coming RSN additional functions in the
mcmc contributed package for R.
- Long range search for maximum likelihood in exponential families (with Sai Okabayashi) Electronic Journal of Statistics, 6, 123–147 (2012).
- On January 20, 2012 I gave the SST (Science and Technology Studies)
Annual Science Studies Symposium. Here is the abstract. Here are the slides
- Departmental Seminar on Parallel Computing.
This talk is seriously out of date. A much better introduction to parallel
computing is my lecture notes for Stat 8054 Spring 2020.
- JSM talk overheads (PDF).
Talk was titled
MCMC: Does it work? How can we tell?. It was
in an invited session organized by
Peter Hoff in
which speakers were supposed to be critical as opposed to claiming some
idea of theirs was the greatest thing since sliced bread like every other
speaker does. I tried.
- Likelihood Inference
in Exponential Families and Directions of Recession (UMn TR 672)
- My Ph. D. Thesis
University of Washington 1990.
- Radically Elementary Probability and Statistics
(UMn TR 657)
- Aster Models Preprints for two papers that
have now appeared, also numerous technical reports, web pages and
overheads for a short course, overheads for several talks.
- Le Cam Made Simple: Asymptotics of Maximum
Likelihood without the LLN or CLT or Sample Size Going to Infinity
(UMn TR 643)
- Computing the Joint Range of a Set of Expectations (with Radu Lazar and Glen Meeden)
- A paper Markov chain Monte Carlo maximum likelihood from
the 1991 Interface Proceedings that is hard to find (not on the
web apparently) and has the first mention of what is now called
parallel tempering in the MCMC community.
is the original form as it appeared in the Interface proceedings.
This PDF form
is just the PostScript form run through ps2pdf without any attempt to
use PDF-friendly fonts and may not look very good.
- Some rants about Markov chain Monte Carlo.
Some fooling around with Java applets.