Monday, June 30, 2014

Granger

There really is no danger
That we'll forget Clive Granger
Nobel prize winner that he became
Cointegrated series brought him to fame
More care in studying trends
Was what he taught us
And gave the concept to his name
From Nottingham to UCSD
With Robert Engle by his side
For econometricians everywhere
He is a source of pride.

Backing Bayes

According to Thomas Bayes
Everything is conditional
And of probability he says
We should not be traditional
When new information comes
You must redo all your sums
With prior and posterior
This approach is superior
Must more realistic
To give you a statistic.

Friday, April 05, 2013

Econometrics Journal special session online

Webcasts of some of the keynote lectures at the 2013 Royal Economic Society conference, held at Royal Holloway University of London, are now available online.

This includes an Econometrics Journal special session on Heterogeneity, chaired by Richard Smith of Cambridge University.

Tuesday, November 06, 2012

Multicollinearity madness

My independent variables are highly correlated
And coefficient standard errors consequently are inflated
The estimates' significance are downright understated
These results are making me totally frustrated

The R squared is high
But the t stats are all low
The reason for this we all know
It's multicollinearity madness
I've got multicollinearity madness

I guess I'll just have to wait 'til later
See if I can get a bit more data

Factor analysis is not my style
And ridge regressions are not worthwhile
There are no restrictions that I can use
Or regressors in the model that I could lose

The VIF is far too high
The tolerance is far to low
The reason for this we all know
It's multicollinearity madness
I've got multicollinearity madness

I guess I'll just have to wait 'til later
See if I can get a bit more data

Saturday, September 08, 2012

XtransX to the minus one X transpose Y


X trans X to the minus one
X transpose Y
I said X trans X to the minus one
X transpose Y
you know that beta hat
is calculated like that

they call it OLS
when you want to regress
it's a fitting tool
that will give to you
the lowest RSS
yes that's the OLS

X trans X to the minus one
X transpose Y
I said X trans X to the minus one
X transpose Y
you know that beta hat
is calculated like that

Ordinary but rather neat
For ease of use it just can't be beat
A simple formula
and the calculation's complete
And you will find it BLUE
when all the assumptions are true

X trans X to the minus one
X transpose Y
I said X trans X to the minus one
X transpose Y
you know that beta hat
is calculated like that

they call it OLS
when you want to regress
it's a fitting tool
that will give to you
the lowest RSS
yes that's the OLS

Saturday, July 14, 2012

Keep to the beta - more punometrics

In this subject make sure
That you don't lose the plot
It will give you some clues,
Likely as not
Make sure you're aware
Of what causes what
For modelling it well
Surely matters a lot.

There are many symbols around
Alphas, gammas and betas
Deltas and rhos, and pi's, chi's and thetas
Upper and lower, two cases of sigma
To sum up it's hard, a right old enigma.

You've got hats, you've got bars
And one or two stars
You've got F's, you've got t's
And hypotheses
Reject or accept
You will need to decide
Both tails to be used
Or only one side?

You need to regress
So you can progress
Good results will impress
And give you success

So keep to the beta
It will make your life sweeter
For nothing is better
Than this famous Greek letter.

Monday, April 02, 2012

Hal White RIP

I am feeling particularly sad today as I heard this morning that Hal White, Professor Halbert White Jr of UCSD, passed away last Saturday morning.  James Hamilton has written a brief appreciation about him  here.

I had the privilege of meeting Hal at the Clive Granger Memorial Conference back in May 2010. I can only concur with what others have said and written about him - he was both an extremely good econometrician and a very friendly and delightful man.

I shall be singing Heteroskedasticity Blues in memory of him later today.



Sunday, March 25, 2012

You've got to look out for the lags

Another attempt at an econometrics poem - this one to end my lecture introducing distributed lags and dynamic models


You asked me for some comments
On the models you have tried
Some instantaneous feedback
On what you've specified

My immediate conclusion
Is you should have tried some lags
Your perspective's far too static
You've got to look out for the lags

You've got to look out for the lags
You've got to look out for the lags
Don't want to be over dramatic
But you've got to look out for the lags

Yes you've made a big assumption
That all adjustments will be fast
Come in the current period
Nothing left there from the past

But responses may be spread out
So think about some lags
Your perspective's far too static
You've got to look out for the lags

You've got to look out for the lags
You've got to look out for the lags
Guess I must be quite emphatic
You've got to look out for the lags

When expectations are allowed for
And agents plan ahead
Adaptively or rational
It's just as I have said

Disequilibrium adjustments
Error feedback to include
Your perspective's far too static
You've got to look out for the lags

You've got to look out for the lags
You've got to look out for the lags
More dynamic than instamatic
You've got to look out for the lags

With general to specific
Please test rather than impose
The restrictions you consider
Appropriate to propose

When using time series data
You have to try some lags
Just be a bit pragmatic
And look out for the lags

You've got to look out for the lags
You've got to look out for the lags
If you do I'll be ecstatic
You've got to look out for the lags