The weather in Goulburn for these 4-monthly meetings always seems to be extreme, and for the meeting on 26 November it was blowing a gale all day. Whatever the weather though, the quality of talks is always high, and here are some brief notes on the most recent offerings.
Janice Scealy spoke on “A directional mixed effects model for compositional expenditure data”. Even though the example she chose was about as simple as they come – three categories of household expenditure – it was a complex journey she took us on through square root transformations and Kent distributions to final parameter estimates on the raw scale.
James Brown asked “Can we use the approaches of ecological inference to learn about the potential for dependence bias in dual-system estimation? An application to cancer registration data.” This nice piece of work-in-progress concerned the use of multiple lists to estimate population size when key pieces of information were still unavailable.
Chong You of the University of Wollongong spoke on “Variational Bayes estimation and model selection for linear regression using spike and slab priors”. Despite my recent foray into Bayesian methods for a course I taught to Graduate Certificate students, this was the talk on least familiar ground to me.
Pavel Krivitsky also of the University of Wollongong spoke on “Modelling of dynamic networks based on egocentrically-sampled data”. He has spoken about aspects of this type of data several times and I think I’m starting to get the hang of it. The main application is to the study of sexual partnerships in human populations.
Bendix Carstensen, Senior Statistician at the Steno Diabetes Center, Copenhagen, Denmark, gave the Knibbs lecture to the Canberra Branch of the Statistical Society on 24 November. He is the spitting image of English actor Daniel Craig, but his line of work is somewhat different! His main suggestion in this talk was to use models where the quantitative nature of time is used explicitly in modeling of occurrence rates, using parametric functions to describe time-effects. This allowed for simple derivation of demographic quantities such as residual life time, years of life lost or years spent with disease. Bendix described the philosophy of the Lexis diagram for representation of multistate data on multiple time scales implemented in the Epi package for R. He drew his examples from the clinical literature around diabetes and cancer, and advocated strongly for more graphs and fewer tables in epidemiological journal articles.
The third speaker at the Health Research Institute Networking Forum on 24 November was Milica Muminovic. She’s an architect and urban designer, interested in “place” and cities. She started off by speaking about her PhD thesis on measuring the non-measurable in terms of the character of a public space. She’s studied cities as diverse as Tokyo in Japan and Split in Croatia. She has now turned her attention to Canberra and the public spaces that she admits are under-utilised even though they may display great design principles.
The second speaker at the Health Research Institute Networking Forum on 24 November was Jackson Thomas. He tackled the ticklish topic of scabies, a parasite causing all sorts of rashes and sores. WHO listed it as a neglected tropical disease in 2013. Current treatments such as ivermectin are not easy to administer, especially in Indigenous communities, and resistance of the mite itself is also developing. Jackson will be studying whether tea tree oil will work, starting off with an animal trial in pigs.
Associate Professor Jennie Scarvell gave this talk to the Health Research Institute Networking Forum on 24 November. That famous sporting injury, the ACL, is the catalyst for her study of osteoarthritis. She had some great videos of bone movement when kneeling, some interesting results and as always, more research questions. The takeaway message was that it’s not true that physical activity causes osteoarthritis!
Wahyu Sutiyono presented the School of Marketing seminar on 11 November. Her co-authors were Ali Quazi, Raechel Johns and Kartika Susilowati (State Polytechnic of Malang, Indonesia). They used document analysis to scan the CSR landscape in Indonesia and compare it to the two-dimensional model of Quazi and O’Brien. The recent democratic reform in Indonesia makes this a particularly fruitful line of research.
I liked the six dot points given in this blog for smartening up your teaching.
A – Awaken the Intrigue—We talk about sometimes not saying anything and creating an experience to really capture their attention.
B – Begin and End Often—We emphasize the importance of unpredictable routines and planned transitions so that kids feel safe but excited about what we’re learning.
C – Create Lots of Contrast—Sometimes we know to say it louder, we need a catchy series title, or a counter-intuitive Bottom Line or even visuals on the screen.
D – Draw them in with Stories—We’re wired for authentic stories, which is why we encourage leaders to personalize and share in appropriate ways every week.
E – Emotion Drives Attention—We know that kids are drawn in when they can empathize and relate, in addition to seeing how it applies in their own lives.
F – Focus on the Big Idea—This may be the most important point in an age of constant information. Say less so you can say it clearer and say what matters most. We want to really impress this on young minds and recycle key truths so they stick.
These cushions are truly cute and would make a memorable lecture on probability distributions for sure! The original post about them is here. They are:
- Baby Blue t Distribution
- Light Pink Log Normal Distribution
- Lilac Continuous Uniform Distribution
- Tan Weibull Distribution
- Olive Green Cauchy Distribution
- Slate Blue Poisson Distribution
- Maroon Gumbel Distribution
- Gray Erlang Distribution
Henry Wolfe of the University of Otago certainly knows how to put the wind up a bunch of students and staff! His talk on 6 November about threats to mobile phone security covered the whole range from signal jammers and fake mobile phone towers to how to turn your phone into a bugging device. On the defence side, he showed us all how to find the International Mobile Equipment Identity of your phone (look under the battery or text *#06#) and offered sound advice on turning off tool such as Bleutooth and Near Field Communication when not in use. On the other hand I think the jury is still out on the relationship between mobile phone use and brain cancer – some weekend reading perhaps!
On 6 November there was quite a festival of presentations from Honours students in Forensic Science here at UC. I attended Elaine Cheung’s presentation, first thing in the morning.
She had carried out a very nice investigation into three classifiers – Bayesian, genetic distance, and multinomial logistic regression – for identifying biogeographical ancestry from SNP data. She manipulated the degree of missing data and the number of SNPs analysed to complete her experiment. Her ROC curves were in many cases very close to perfect, and she made the interesting decision to swap to reporting area OVER the curve instead of UNDER the curve to highlight the results of interest. I think there’s another whole Honours project in Stats, about the behaviour of the area OVER the ROC curve!