Heteroskedasticity: Friend or Foe?

[Poker Cat]

Can someone explain homoskedasticity in layman's terms?

I think I kinda know what it means, and thinking it might be applicable to some PT samples.

I studied it my last year in college (um, 1985). Now that it might actually be useful, I have to learn it over again.

[Sharkey]

The tightness of the data points around the regression line doesn’t vary with x.

How are you using it?

[Poker Cat]

I didn't want to do this, in case my 21-year-old recollection is dead wrong, but . . .

I thought homoskedasticity to be a measure of how well smaller, random sub-samples match the features of a larger sample that contains them. In others words, consistency.

Here's how I thought it might apply to poker. If you have a fairly large sample of hands, say 200K -- that's large enough to be fairly confident of your win rate, but not large enough to be really accurate. I was wondering if you could take random sub-samples of those hands -- say 20K each, and do a Goldfield-Quandt test to see if your win rate was consistent enough to increase confidence in the accuracy of the larger sample.

[Sharkey]

[ QUOTE ]
how well smaller, random sub-samples match the features of a larger sample that contains them.

[/ QUOTE ]

What you’re describing is a function of “variance”.

[ncray]

[ QUOTE ]
The tightness of the data points around the regression line doesn’t vary with x.

How are you using it?

[/ QUOTE ]

Heteroskedasticity means that the variance of the conditional distribution DOES depend on x. Homoskedasticity means that it's constant variance over all x.

[Sharkey]

The question in the OP was about [censored].