- PhD in statistics, from QUT & Paris-Dauphine
- Honours in Bioinformatics (Griffith), BSc in Statistics (Otago)
- I now live in Brisbane, by way of a few places
The most common question asked since I started to pursue Statistics has been:
I can share my top three reasons!
The exponential growth of computing is not slowing down!
It is notoriously hard for our brain to really comprehend what this means.
Opinions are changing fast, and everyone is coming on-board!
Amazing things happen when data analysis combines clear research questions, appropriate data, and suitable, accessible tools.
Alzheimer's Disease (AD) currently affects over 342,800 Australians, and this number is expected to rise to 900,000 by 2050.
Cognitive changes occur very late in the disease (\(\geq\) 20 years).
During this time, AD causes irreversible damage to the brain!
SUVR available for 393 individuals: 290 HC, and 103 AD
Originally, compared AD to HC, and so on…
But something quite interesting is happening here.
We expect some HC's have early stage AD.
This means the HC data must contain a mixture of individuals, HC and not.
But if different subgroups exist, can't compare AD to HC!
Undetected subgroups can cause problems…
We have a mixture distribution with an unknown number of groups.
Traditionally, these can be quite painful!
Install the package
devtools::install_github('zoevanhavre/Zmix') # Thank you Hadley! library(Zmix)
Run the model (with 5 groups)
Zmix.Y<-Zmix_univ_tempered (Y, iter=50000, k=5)
Process the results