Any mathematicians in?

[Alex]

This is, admittedly, a very long shot, but here goes...

I'm trying to understand a medical journal about the relationship of obesity and risk of death. The relationships between quintiles of waist circumeference and death were analyzed using non-parametric restricted cubic splines.

Now I've been googling for hours, and reading a few basic textbooks for a while, and I'm more confused than when I began.

Does anyone in the world actually know what a non-parametric restricted cubic spline is? And care to explain it to me in understandable terms?

I thought I was quite good with statistics until I came across this!

[Fraz]

n00b.










Ahem, more seriously, I thought I was quite good at maths also. I've heard of splines (common in computer graphics), but I certainly have no idea what a non-parametric restricted cubic spline is I'm afraid.

At a total and probably incorrect guess, I would say they are using the aforementioned splines as a way of modeling the arbitrary function that relates waist circumference to risk of death.

[Steve]

The fatter you get, the more you die.

[Fraz]

The fatter you get, the more you die.

Presumably the thinner you get the more you die too, below a certain point.

[MadduckUK]

i tried to break it down into the separate terms to figure it out and now i have a nosebleed cheers for that

[Whiternoise]

"In the mathematical field of numerical analysis, a spline is a special function defined piecewise by polynomials."

Ok.. so for each range of values it takes a different polynomial function - so between 0-1 you might say it's "X^2" but between 1-6 it's "X^3"? As wikipedia says, the simplest spline is a step function - something you might see if you're working with logic i guess. But it sounds like it's basically a curve made up of lots of individual "curve" pieces. Which makes sense for data that doesn't fit any proper line of best fit. You just make a spline that fits the data?



That is apparently a cubic spline.

"Non-parametric methods are widely used for studying populations that take on a ranked order (such as movie reviews receiving one to four stars). The use of non-parametric methods may be necessary when data has a ranking but no clear numerical interpretation, such as when assessing preferences."

That would explain the non parametric bit no?

And a restricted cubic spline is a "natural" spline. "The word "natural" means that the second derivatives of the spline polynomials are set equal to zero at the endpoints of the interval of interpolation". Which would mean that the rate of change, of the rate of change of the function is constant - however that helps..

http://www.cse.unsw.edu.au/~lambert/.../natcubic.html
http://en.wikipedia.org/wiki/Spline_..._interpolation

But that's as far as i get.. i think it's just a good way to approximate a function that can't really be described by standard regression techniques - like fitting e^x, x^n type curves over the data.

[Alex]

The fatter you get, the more you die.

LOL, basically. Though apparently the ideal BMI for men is 25.3, which is techincally overweight. So you're more likely to die (nice way of putting it) if your BMI is 21-25 (ideal weight) than if it's 25-28 (mildly overweight). Also more likely to die if your BMI is less than 21, i.e. underweight.

And for some reason a huge group of American scientists just finished studying 519000 europeans to find this out. And then explain it in non-understandable language.

Go figure.

[Steve]

Presumably the thinner you get the more you die too, below a certain point.

No, it's not not inversely proportional. People with zero mass live forever.

[Fraz]

LOL, basically. Though apparently the ideal BMI for men is 25.3, which is techincally overweight. So you're more likely to die (nice way of putting it) if your BMI is 21-25 (ideal weight) than if it's 25-28 (mildly overweight). Also more likely to die if your BMI is less than 21, i.e. underweight.

And for some reason a huge group of American scientists just finished studying 519000 europeans to find this out. And then explain it in non-understandable language.

Go figure.

Surely all you really need to look at is the result plot, and maybe the equation where you enter your waist circumference and whatever other parameters and it outputs your risk of death, right? If the paper is lacking these things then I suggest you email the authors and tell them they are a bunch of useless rubbishrubbishrubbishrubbishrubbish.

[Whiternoise]

http://content.nejm.org/cgi/content/full/359/20/2105 that one by any chance?

EDIT: I dunno if it's sign in only though, i can see it through the uni network (quite fun looking at all the mad journals for free ).

[Fraz]

No, it's not not inversely proportional. People with zero mass live forever.

True true. At zero mass, you'd become luminal. Time would dilate to infinity, and you'd live forever.

A rather ironic side effect of being so thin is that you'd also be everywhere at once - spacial dimensions would all contract to zero length from your frame of reference.

[Saracen]

The fatter you get, the more you die.

I'd better watch what I eat. After all, I wouldn't want to die too much.

[shadowmaster]

The fatter you get, the more you die.

Lol, I wish medical journals were as simple as that.

[Zak33]

cubic equation
A polynomial equation of degree 3.

Start there.... then sink without trace as I did

[Whiternoise]

Ah maths is easy..!

What do you want to know zakky?

[Yosh]

True true. At zero mass, you'd become luminal. Time would dilate to infinity, and you'd live forever.

A rather ironic side effect of being so thin is that you'd also be everywhere at once - spacial dimensions would all contract to zero length from your frame of reference.

I come in peace! Beam me up Scotty