Google


Friday, January 25, 2008

The end of GLE


The fraud case by a single trade is a heavy blow for Societe Generale. It is of course quetionable how he could conceal his actions.

It's share price has already dropped by half from May 07 to now. Will more equity be destroyed? Will it be overtaken?

Sunday, January 20, 2008

Pay to become Mayor

Why do we have to pay 10 grand to stand in the London Mayor election? How undemocratic. You only get them back if you get above 5% of the vote, so big parties with donors are favoured.

Picking the unripe cherries

Germany has seen some acceleration in its growth recently, after some hard reforms. Now we see the groups who had to hold back for some years reemerging from their trenches and holding the hands open. Trade unions and SPD are asking for more money. Beck is telling people not to buy Nokia mobiles because they leave Germany - ridiculous. They undo the hard won success of their predecessors.

Germany might still be Exportweltmeister, but still they will be overtaken by the Chinese and then the Germans won't be top at anything.

Data loss

The frequency of known data loss by the government is astonishing and shocking. What about all the cases we don't know. One problem is that many civil servants work from home and the IT is exposed. Another problem is the lack of security systems or better the lack of usage of these systems. It is pretty easy to decode something, and people have to be trained and reminded of the importance of data security.

Wednesday, January 16, 2008

LSTAR

Here is a SAS macro to estimate a logistic STAR(1) model, the first part initialises the parameters by using the OLS coefficients, this avoids the optimisation going off in the wrong direction.

options mprint mlogic;

%macro star(data=,var=,gamma=10,c=);

%if &c= %then %do;
proc summary data=&data;
var &var;
output out=temp;
run;

data _null_;
set temp;
if _stat_='MEAN' then call symput('c',&var);
run;
%end;

data temp2;
set &data;
l&var=lag(&var);
ll&var=lag(l&var);
run;

proc reg data=temp2 outest=temp3;
model &var = l&var ll&var;
run;

data _null_;
set temp3;
call symput('a',intercept);
call symput('b',l&var);
call symput('d',ll&var);
run;

* y_t = phi00+phi01*y_t-1 + (phi10+phi11*y_t-1)*(1/(1+exp(-gamma*(y_{t-1}-c))));
proc nlin data=temp2 maxiter=500 convergence=0.000001;
parms a0=&a b0=&b c0=&d a1=&a b1=&b c1=&d gam=&gamma c=&c;
model &var = a0+b0*l&var +c0*ll&var+ (a1+b1*l&var+c1*ll&var)*(1/(1+exp(-gam*(l&var-c))));
output out=temp4 predicted=&var.f;
quit;
%mend;

* Usage;
%star(data=unemp,var=gmgsx,gamma=30);

Monday, January 07, 2008

Davos 2008

From 23 to 27 January the World Economic Forum takes place in Davos. Co-chairs will be among others Tony Blair and Henry Kissinger. One of the main questions will be if the global economic slowdown can be halted or reversed.

Friday, January 04, 2008

Fuel consumption


When environmentalists talk about energy consumption they often talk in absolute terms. From 1971 to 2005 world fuel consumption increased by 88% from 4.2 billion tons of oil equivalent to 7.9. In the same time world population increased by 76%. This implies that fuel consumption per head only increased by 7% over 34 years.