Re: (statistics)how to make date more like Laplacian distribution?

In article <[EMAIL PROTECTED]>,
 "walala" <[EMAIL PROTECTED]> wrote:

> Why people think I want to lie upon seeing my question? Oh, it's my problem
> that I did not clearly present the background...

Oh, I'm not calling you a liar. It does always help
that you explain the source of your data. In this
case, I'll define it as more like creative accounting. ;-)

"There are three kinds of lies: lies, damned lies,
and statistics."

       Mark Twain, Autobiography. 


Ok, I realize you have a serious question. I do
not have an explicit answer for you though.


> Here is the story: in deblocking of block DCT coded JPEG images, it was
> known that the DCTed coefficients are Laplacian distributed... But now I am
> looking at low bit rate JPEG images, so there are someking of artifacts...
> in order to reconstruct the original images... many algorithms have been
> devised... one possibility is to make the image coefficients more Laplcian
> like...
> 
> So that came my question: how to make data more Laplican like...

Its a little tough, since I don't know what the
deviation from Laplacian-ness is.

If the empirical distribution of x does not have
the desired shape, one could transform x so that
its moments match the moments of a laplacian.
Since there are many possible transformations,
it is not at all obvious which one would be
appropriate. (I honestly don't know.) You would
need to try out some different ones to see which
works best. This would appear to involve a
nonlinear root-finding operation.

Perhaps you do not appreciate why I cannot say
more. I'll give you an example. Suppose you had
two distinct cases. In the first, instead of a
Laplacian distribution, the distribution of
coefficients was normally distributed. In the
second case, the coefficients turn out to be all
either zero or one, with probability 1/2. You
can appreciate that each case would require a
different class of transformation.

Perhaps others will have better ideas.


> (I am just a poor student, not lieing government agency, issurance company,
> weapon dealer, lawyers, and politicians... so please help me!)

Government agencies never lie. They do not need to.
They just change the rules so that whatever they
say becomes the truth. ;-)

HTH,
John


-- 
There are no questions "?" about my real address.

The best material model of a cat is another, or
preferably the same, cat.
A. Rosenblueth, Philosophy of Science, 1945