Musings

[clawbeast]

I have a few thoughts that I thought I would float out into the aether so that those with greater minds than I might give their opinions

Comments

[ciderclaus]

Ultimately if you don't try and work something out you'll never know if its going to work. How much better will things be if it works against the input of time anmd effort if it doens't. I suggest giving it a go.

[sack_boy]

Need to collect a lot of measures etc for all patients and controls.

Then you're might want to have a look at something like Principle Component Analysis on the outcome.

Or summat like that.

[clawbeast]

I have a test base here with 250,000 anonomised coronary patients on it. As well as quantifiable measures (eg cholesterol, ldl, hdl, fbc, weight etc), other medical issues and family histories.

Already applying PCA and SPC principals to the data, but wondering about a fresh take. Though I should feed a common set of data in, otherwise the conclusion will be people with test x are safe.

[mr_kobayashi]

Don't forget to consider quality of life.

[clawbeast]

General idea is that the NHS would save cash, and the patient gets a better quality of life.

[glenatron]

I don't know whether a neural net would be a good tool for this type of analysis- I've only explored them at a toy level rather than a practical one - but these people think they are. Certainly they're good for pattern recognition, but whether they are better than other statistical methods is another question. Of course, you can find ways to combine them, no question.

I suspect that if you were to start exploring this field you would be treading fairly new ground, which may mean some hard work to get started, but possibly some genuinely innovative outcomes.

In terms of the statistical stuff, have you explored using Bayesian techniques? Those seem to be turning up everywhere at the moment as some kind of statistical swiss army knife...

I find the whole AI area absolutely fascinating, although I know endlessly less about it than I would like to.

[irdm]

"the science of trying to work out what makes people"
Well, I don't know if you've noticed yet, but I'm pretty sure that the answer is "shagging".....

In a more related snippet, 2 of the systems I'm currently working on are for collecting episode data for BSUG & BSGE although I'm fairly sure they don't know how to get from A to B with analysing it....

[clawbeast]

Out of interest what method are you using for capturing episode data?

[irdm]

Consultants type it in (read: consultants get their minion to type it in)

Years ago they had an Access front end which they extracted and emailed off to someone. Which was the original spec.
We wrote a similar UI but on a website so all the data is in one place all the time. No rockets were scientisted in the making of this app.
They've almost finalised what data to collect and one group are actually trying to decide what reports they want from it.

[gopher_that]

What is your starting hypothersis?