Cabaret ISBA 2008

Thursday was the last day of talks at ISBA 2008, culminating with a keynote by Tony O'Hagan on elicitation. Overall this was a great conference. I enjoyed the talks, and felt that they hit the right mix of diversity on topics and level.

Of course, at a Bayesian conference, the end of talks is when the party begins. That night was the banquet, followed by the entertainment. At all the Bayesian conferences I've been too, the entertainment is provided by the attendees at a Cabaret that anyone can join in on. I myself sung a little ditty called "The Simulator" accompanied by Gareth Roberts and Jeff Rosenthal. There is a video up on YouTube for those brave enough to watch--I'm not putting a link because although the crowd was into it, as a performance piece it's not my best work. I do plan to put a recording of myself singing up on my site when allergy season is over. For now, though, enjoy the lyrics, sung to the tune of "The Gambler":

On a warm summer's evening, on a plane bound for ISBA,
I met up with the Simulator, we were both too tired to speak,
so we took turns a'staring, out the window at the darkeness,
'Til boredom overtook us, and he began to speak.

He said, "Son I've made a life, out of reading people's data",
and knowing what their numbers tell, but the way their series lies,
So if you don't mind my saying, I can see you're out of models,
for a look at your data, I'll give you some advice."

So I handed him my laptop, and he downloaded my last file,
Then he bummed a thumbdrive, and I watched its blinking light,
and the night got deathly quiet, and his face lost all expression,
he said, "If you're running chains, boy, you gotta learn to run them right."

You've got to know when to propose them, know when to reject them,
know when to stop a chain, know when to run,
you never find your error, while your taking samples,
there'll be time enough for error bars, when the chain is done.

Now every runner knows, that the secret to chain burn-in,
is knowing what to throw away, and knowing what to keep,
because every run is perfect, and every run is worthless,
and the best that you can hope for is results you can repeat.

And when he'd finished speaking, he turned back towards the window,
Shut down his black ThinkPad, and faded off to sleep,
and somewhere in the darkness, his chain it reached convergence,
But his final words, I found a trick that I could keep.

You've got to know when to propose them, know when to reject them,
know when to stop a chain, know when to run,
you never find your error, while your taking samples,
there'll be time enough for error bars, when the chain is done.

You've got to know when to propose them, know when to reject them,
know when to stop a chain, know when to run,
you never find your error, while your taking samples,
there'll be time enough for error bars, when the chain is done.

You've got to know when to propose them, know when to reject them,
know when to stop a chain, know when to run,
you never find your error, while your taking samples,
there'll be time enough for error bars, when the chain is done.