if you don't understand why this is amazing check the comments
An (American) politcal expert on "emerging democracies" or similar was on the radio, trying to advance the thesis that the Afghan people are eager to have an election, which is apparently supported by polls. He was challenged by the host, who pointed out that the Taliban were part of the Afghan people and they did not support elections.
He replied that a poll had shown that only 4% of people in Afghanistan supported the Taliban.
Exercise: Suggest two ways in which this statistic may be flawed.
I wrote this down in my notebook and no longer know why
***STOP: 0x000001E (0x0000005, 0x80084761, 0x00000001, 0x0000001C)
(MODE_EXCEPTION_NOT_HANDLED
Beginning dump of physical memory
Physical memory dump complete. Contact you system administrator or technical support group.
Plain Omellete & Chips
Cheese Omellettes, Chips
Ham Omelette & Chips
Mushroom Omellete & Chips
Spanish Omelette & Chips
As linked to on the WOWblog
This is interesting, a picture showing how big all the water and air on our planet would be if they were contained in spheres:
http://files.abovetopsecret.com/uploads/ats52865_Globa_lwater_air_spheres.jpg
I was recently looking a David Li's paper 'On Default Correlation: a Copula Function Approach'. This is the paper that has become slightly notorious as containing the 'equation that caused the crash'.
The 'Copula Function Approach' is to consider the joint distribution of two variables by mapping each one to the uniform distribution using its cumulative probability distribution function. That is to say, if X is a random variable and f(x) is the function P(X < x), then F(X) is distributed uniformly on [0,1]. If we do this for 2 random variables X and Y, the joint distribution of F(X) and G(Y), each of which is distributed uniformly, gives us the correlation structure of X and Y, abstracted from their individual distributions.
So we can take any two random variables X and Y along with a copula function H which satisfies certain properties, and create a joint distribution for X and Y with the relevant copula describing their interaction. The transformations from X and Y to the uniform random variables F(X) and G(Y) does not preserve their Pearson correlation coefficient, but does preserve the Spearman coefficient, the rank or non-parametric correlation coefficient.
Li's model takes two individual 'survival functions' - the probability that some party will not have defaulted on their debts by time t, and a Gaussian copula function, and uses the joint distribution he gets from these to estimate the value of eg insuring the credit of one party, where the other party is the underwriter. The 'Gaussian copula function' is the copula one gets from taking X and Y as Gaussians, and their joint distribution as a bivariate Gaussian. The bivariate distribution depends only on 1 parameter: the correlation coefficient between the two Gaussians. So given a correlation coefficient supposed to hold for two assets, and a survival function for each, we can get a bivariate distribution which is supposed to reflect all the combined behaviour.
The interesting part is what Li chooses for the correlation coefficient. He simply takes the correlation coefficient between the prices of the (credit default swaps on) the two assets. As has been pointed out many times, this fails to account for 'systemic risk' or higher correlation at times of crisis. However it seems that apart from this, there are some major problems with this:
1) Even without systemic crisis, correlation between prices and correlation between default times should be similar only at the time of default (when both price and perceived survival probability go to 0, probably pretty closely together). In general there is a lot of noise in price changes that there isn't in the survival function, and therefore prices are less strongly correlated.
2) Even without noisy prices, correlation of prices and correlation of survival functions shouldn't be the same. At best they should be related exactly by some function, which should be able to be calculated from their definitions. There isn't a linear relationship between price and default time - so even if there is a precise, deterministic function which relates the two, the two sets of correlations are not necessarily equal.
that tickets for centre court at wimbledon on saturday are "literally" worth their weight in gold.
however as the tickets only weigh about 5g each this is equivalent to about £90 and is a vast understatement on the literal truth.
Recent Comments