?

Log in

No account? Create an account
A random question for people with science leanings: I need a… - The Mad Schemes of Dr. Tectonic [entries|archive|friends|userinfo]
Beemer

[ userinfo | livejournal userinfo ]
[ archive | journal archive ]

[Sep. 23rd, 2005|01:57 pm]
Beemer
A random question for people with science leanings:

I need a probability distribution, but its shape isn't really known. That inclines me to use a gaussian as a generic, because most things are gaussian.

However, it can't produce negative values. What would you use?

Truncate the gaussian below zero? Fold negative values back to positive? Something like a chi-square distribution that looks gaussian once the mean is far from zero?
LinkReply

Comments:
[User Picture]From: ocschwar
2005-09-23 02:10 pm (UTC)
When you convolve a probability distribution with itself, it converges towards a Guassian. When the distribution is constrained to only have non-zero probability density in the positive half, it looks more like the first derivative of the Gaussian. (Take a trncated Gaussian and convolve it with itself in your mind's eye.)
(Reply) (Thread)
[User Picture]From: melted_snowball
2005-09-23 02:24 pm (UTC)
Your second sentence contradicts the first, no? The central limit theorem applies to any distribution.

[I admit that I'm awkward with the style of language you're using here, but at least that's what I get when I look at what you've written...]
(Reply) (Parent) (Thread)
[User Picture]From: ocschwar
2005-09-23 03:00 pm (UTC)
After some coffee I'll rephrase that, if I can make those who read it not lose IQ points. *sigh* It's been a day.
(Reply) (Parent) (Thread)
[User Picture]From: melted_snowball
2005-09-23 03:07 pm (UTC)
Yes, it has. I hate teaching about finite automata...
(Reply) (Parent) (Thread)