Math n00b Question: Share your skillz please, Tease Your Brainz

I don't know what the formula is, but statistically it's worthless. Variance needs all the data points, not just the high and low.

All that doesn't mean some group doesn't use it, but if that is the financial industry it's no wonder they screwed up risk assessment so badly.

It might help of you set it equal to something. It's a polynomial BTW.

Often times these are set equal to one.

The formula is calculating a percentage with known numbers, not solving for them. Setting that equation equal to anything is incorrect and even less meaningful than the percentage is.

No. You are misapplying what the math of statistics means and does. The screw-ups most folks see in situations which cite statistics is almost entirely due to fatal oversimplification of those computing and/or interpreting the results.

If you aren't interested in actually getting the subtleties of the calculations correct it is best to run away from any form of mathematical analysis and just guess. At least you will be honest with yourself that way. In a way I think none of the financial houses were.

See those statistics do mean more than bollocks. When an idiot uses the information utterly incorrectly it has no reflection on the statistical analysis. You are even showing serious ignorance in your statements about bell curves, properly called normal distributions. See anyone who actually respects their ability to do analysis knows the tails of those curves go out to infinity entailing all possible outliers. They also understand that there are an infinite number of ridiculous far outliers out there.

I agree that the financial models need an overhaul. Unfortunately nobody knows enough about the specifics of human nature to make those models. It was the hubris of a few programmers, fed by a few years of up market momentum, saying they had a good model because they refused to acknowledge they didn't have the first clue about many critical variables. Or even worse, they didn't have a clue that they didn't have a clue.

If you want a formula for variance, just go for the basics: Discrete_case

Well then do this

((x-y)/((x+y)/2))*100=p

or even this

((x-y)/((x+y)/2))*100=f(x,y)

Interestingly enough, the same formula is used for "modulation", which describes the relationship between resolution and sharpness of an image. It is approximately the same as "contrast".

Which makes sense somehow, because nvidia2008 wants to know the "contrast" of the exchange rates.

Unrelated to this thread, this morning I found a paper from ARRI (those guys who build expensive cinema camera equipment) where this formula and its relevance is explained on page 7:

Theory Basics for Motion Picture Imaging

NVidia, you keep saying "maths" instead of "math". Would I be right in assuming you have a non-North-American background? The rest of the world abbreviates mathematics as maths.

Cheers

I think they are part of the root cause and responsible for much of the volatility.

You're welcome.

Of course the ARRI guys recommend using 35mm, they make the cameras... Film may still be the better way for the next couple of years or so, but prohibitively expensive (for my work, anyway). We will see a few quite interesting things in the next year, i. e. Red Epic/Scarlet...

OK we have a name for the function, modulation transfer function. It describes a rule of how to interpret data so I guess I was a bit strong saying it is statistically worthless.

But, from a predictive statistics standpoint it is still quite worthless.

That paper, which I don't dispute in the least, is doing fixed mathematics on fixed data to find a closed form solution in a closed system. That is the right approach for what the whitepaper is discussing. We have a simple comparison which is very useful for comparing discrete points in a fixed image (the modulation transfer function) which helps determine the point at which the data resolution is too small to discern any remaining differences in the data.

Taking an approach out of it's domain takes the mathematics out of context and can lead to all kinds of badness because the baked in assumptions which are necessary for correctness no longer apply. Sometimes creative folks come up with ways to address the issues and show that the equations are more general and that drives the overall understanding of the world forward, but this isn't what we have here. This says nothing predictive about the trajectory motion of the data through time.

Probably early next year. If so, I keep you updated.