ASC 2014 poster session

I was delighted to present a poster at The Australian Statistics Conference, ASC 2014, with a former student of mine, Bo Cui. His work on “Australian teachers’ intent to leave teaching profession through logistic regression analysis” formed part of a project unit in the Graduate Diploma of Statistics offered at the University of Canberra. Well done Bo!

ASC 2014 Bo poster 02 compress

Comments Off

Pi Day

Hapy Pi Day (22/7) everyone! Here’s the batch of pi-kelets I made for the Pi Day event at the University today.

image

Comments Off

ASC 2014: Keynote speakers

I’ve got a series of posts now about the Australian Statistics Conference, held in Sydney from 6 – 10 July. Firstly, the two keynote speakers that I heard.

Terry Lyons, University of Oxford

His example was writing the précis of a text. Stochastic differential equations got an airing as did the miraculous shuffle product.

James Brown, UTS, “Measurement of” and “adjustment for” census coverage
James presented the 2014 Ken Foreman Lecture on this topic. His wide-ranging talk on the UK 2001 and 2011 censuses included the notion of dual-system estimation, like capture-recapture estimation. His funniest comment was the way imputation into the census database could provide husbands for likely-looking ladies so that Bridget Jones need not have worried so much as she did in the movie!

Comments Off

Building capacity in statistics majors: drawing strength from a diverse region

I used the ESTeM Friday 4 July seminar spot as an opportunity to practise the talk I’ll be presenting at ICOTS in two weeks’ time. A cold day in the winter term did nothing to encourage quantity of audience but I was very happy with the quality. In this talk I drew on my experiences in Vietnam last year to put together experiences teaching statistics around the Pacific Rim. Then I tried to draw together the recommendations from the Vietnam conference into a coherent set of goals that have worldwide relevance.

Comments Off

Statistics in action part two

Twenty-one chapters about the Canadian contribution to statistical theory and practice. I thought I’d try writing pi-ems about each chapter, where the number of words in each line of the poem is 3, 1, 4, 1, 5, 9. To avoid the posts getting too long, I’ll do them on batches of seven. Here’s the second instalment!

Chapter 8. Statistical genetic modelling and analysis of complex traits.
Simple odds ratios,
Bayes,
Enough to get started
On
GWAS, NGS and the like.
Finishing touches on gene-environment interaction and sample size.

Chapter 9. Bayesian methods in Fisher’s statistical genetics world.
The basic problems
(Regression)
Are simple. But with
400,000
hypothesis tests and winner’s curse,
Polite comparison with frequentist methods shows Bayes is key.

10. Bayesian methodology for observational health sciences.
Epidemiology and Bayes
intertwined,
holy trinity of threats
avoided.
Choosing difficult not big data,
I like this author’s style, Canadian at the end.

11. Statistical models for disease processes: markers and skeletal complications in cancer metastatic to bone.
Funded by Canada
Using
Patients from Canada (probably)
But
Very generalisable to other locations.
Five-state biomarkers, Kaplan-Meier curves and hazard rates.

12. Analysis of biased survival data: the Canadian Study of Health and Aging and beyond.
Alzheimers, dementia, aging,
Survival.
A highly personal account
Of
Ups and downs in research.
New data should shed light on this important topic.

13. Assessing the effect on survival of kidney transplantation with higher-risk donor kidneys.
Analyse effects of
Kidneys
Not ideal but transplanted
Nonetheless.
Sequential stratification is a solution,
Focusing on the effect-of-treatment-among-the-treated.

14. Risk-adjusted monitoring of outcomes in health care.
An intuitive concept
Thrice.
VLAD (not the impaler)
EWMA
RA-CUSUM completes the set
With a nod towards Australian work at the end.

Comments Off

Annual review of statistics

This new journal was launched recently, with the first issue downloadable for free. What I was hoping for here was something akin to the astronomy annual reviews (published on DVD, I suppose astronomy is a bit more visual than statistics … Sometimes …). I would read 22 articles about the latest developments in statistical practice in a variety of application areas. What I actually got was 22 articles much more about the state of the art of statistical practice in a variety of application areas. One paper even reported on a new experiment to test the effectiveness of various models for observational data. I suppose when you publish the first annual review you get a chance to survey the entire history of a field … But these were not all history papers either. The level of theoretical treatment is slightly lower than, say, the Canadian volume “Statistics in Action” (see elsewhere in this blog).

Comments Off

The effect of environmental factors on fabrics and fibres

This was an student Honours seminar in the Forensic Studies Program at UC on 25 June. Luisa had conducted a nice little experiment to study the effect of several factors on colour loss in fabric samples. The rather large number of factors compare to the number of samples probably prevented a full analysis of variance taking place, and the colour loss is measured in two ways. One is a some scale from 0 to 8, the other involves a spectrometer analysis, much more complex, Her statistical methods for this output included principal components analysis, resulting in a strongly patterned plot of the first two components. I couldn’t quite get to the bottom of why the plot was so structured, and I’m looking forward to ongoing discussions with the forensic scientists about their data and clever ways to analyse it.

Comments Off

Taming non-response in finite population sampling using auxiliary information

Dr Siu-Ming Tam of the ABS gave this talk to the Canberra Branch of the Statistical Society on 24 June. It was a reworked version of a talk he gave in Goulburn not long ago (not that I remembered! And I don’t think there’s any harm in hearing the same material multiple times). The slides were heavy in equations, and I wondered if there are any visual representations of them that could provide the audience with multiple ways of understanding the concepts. A shorter version of this talk will be given at ASC next month.

Comments Off

Neural knitworks

I’m going to take part in this collaborative artwork for Science Week. I’ll be knitting as many luridly-coloured neurons as I can before mid-July, to contribute to the travelling exhibition of a woolly brain. What a great combination of science and art!

http://www.scienceweek.net.au/neural-knitworks/

Comments Off

Enhancing laboratory diagnosis of infectious diseases using modern statistical methods.

I gave this talk to the CReSTS research group in our faculty on June 6, following up on my main study leave report to the Faculty at large. I spoke mostly about hepatitis, concentrating on published works by Brett and myself, Guifang and others including myself, as well as the unpublished paper Brett, Tony and I are still revising. The audience comprised mostly PhD students along with staff from ESteM and Health. The small size turned out to be a bonus, allowing me time to explain methods like decisions trees and plant the seed of the idea of the usefulness of modern statistical methods in a few key heads.

Comments Off