Thanks for the mention Madelyn. It's me who is building the team. The speed with which the message has got around just goes to show the power of the internet.
The UK government is aware of the latent power of raw data. It has an Open Data initiative to get as much as possible into the public domain. Details can be found on their web site:
http://data.gov.uk
An example of the data that can be accessed is a database containing for each of around 10,000 NHS GP practices in England the number of prescriptions they write for each drug for each month.
But this is just the tip of the iceberg: the real power comes from connecting data from one database with data from another database. For instance, as I've posted here previously, I've combined the prescription data with location data to construct maps of the inferred prevalence of Parkinson's across England. For an update on my post see:
http://www.parkinsonsmeasurement.org...ceEngland1.htm
The government site has a forum, part of which is devoted to building teams to tackle particular problems. (Something worth considering for our forum?) The volume of the data is far too much for one person to data mine exhaustively. Hence, the post that Madelyn reports.
Moving forward, as I see it, we have a huge problem, Parkinson's, but we also have:
- a huge amount of data which, I believe, contains the answer as to what causes Parkinson's;
- a group of people with a huge range of skills;
- a huge amount of computer power, which sits idle most of the time.
Put all these things together properly and we can make a huge difference.
I think we can best work together while working apart, by sharing our time, our tools, our methods, our data and our results. That's what this forum does better than any other I've found. Can we ramp this up to account for the huge opportunity that this vast amount of data gives us?
As a start, please let me make a proposal. In the next 5 DAYS (that's 5 days, NOT the usual 5 years) we each find the correlation of one thing with Parkinson's. And let's set a team target of 42 correlations.
The important thing is to get started, so don't, at this stage, worry about the mathematical preciseness of correlation - though we'll need to treat it rigorously later. Don't worry about whether the correlation is positive or negative or zero, all results are interesting. And, although correlation is not the same as causality, don't worry about that for now; we'll deal with that later.
How to get started? Go with what you're good at and use what data you have access to. Use literature surveys, excel analysis, comparisons of geographical distributions. As for data, there's the UK government site already mentioned, in the US there's wonder.cdc.gov, there's the MJFF backed Parkinson's Progression Marker's Initiative,
www.ppmi-info.org etc..
Think inside the box. Knowing what's known is worth knowing. Parkinson's and gender, and age, and smoking, and milk ...
Think against the box. Does the data support what's "known"?
Think outside the box: if it was obvious, someone would have found it by now.
If you're stumped for what to do next:
- find the prevalence map of Parkinson's in the US, reported by Willis et al. in:
http://www.ncbi.nlm.nih.gov/pmc/articles/PMC2865395/
- choose something that comes to mind, anything that may vary across the US, and google
x distribution US map
- choose the Images filter (not Maps, maps will be included in the images.)
- find maps that look interesting, compare these with Willis's results.
- report back.
Remember, 42 by Thursday!
John