RE: [sv-ec] dividing values among bins

Shalom, 

 

This is precisely the type of problem that we avoided with the existing
simple rule. While an even distribution of bins may be valuable, it's
important to note that not only does the rule need to be unambiguous,
but it must also be simple enough for users to figure out what happened.
Rules that change the rounding mode are not the kind of thing a users
can quickly (and correctly) do in their heads. In practice, the current
mechanism works well because the auto-distribution feature is often used
when the bins are a multiple of the range (i.e., users know it), and
when users really care about the bin distribution, they can forgo the
short-hand and specify them explicitly.

 

      Arturo

 

________________________________

From: owner-sv-ec@eda.org [mailto:owner-sv-ec@eda.org] On Behalf Of
Bresticker, Shalom
Sent: Monday, February 12, 2007 7:19 AM
To: Dwivedi, Swapnil; SV_EC List
Subject: RE: [sv-ec] dividing values among bins

 

For bins "b[100] = {[1:401]};", you could distribute them either as
4,4,4,...,5 or as 5,4,4,4,....4.

It does not matter whether the extra values are distributed among the
first bins or the last bins.

It does matter that they not be all concentrated in one bin.

So if (# values)/(# bins) = N with a remainder of M, then either the
first M bins should get N+1 values and the rest should get N values, or
the last M bins should get N+1 values and the rest should get N values.

I do think it should be consistent whether the extra values are in the
first bins or in the last bins.

Shalom

 

________________________________

From: Dwivedi, Swapnil [mailto:swapnil_dwivedi@mentor.com] 
Sent: Sunday, February 11, 2007 10:35 PM
To: Bresticker, Shalom; SV_EC List
Subject: RE: [sv-ec] dividing values among bins

 

Hi Shalom,

 

Consider your example

 

bins b[100] = {[1:401]};

 

With your rule, the case of "401" values distributed amongst 100 bins
would need special definition. I.e the most logical distribution would
be

 

4,4,......4, 5

 

I think you do have a valid point and concern. I might second your
proposal if it goes mid-range and suggests using round-up (or round
-down). Basically, for your example, if values per bin is 4.5 and above
then use 5 values per bin for first 99 bins and rest in last bin. For
lesser than 4.5 values per bin, use 4 values per bin for first 99 bins
and rest in last bin. The standard deviation of difference between
number of values in first 99 bins and number of values in last bin would
be lower with this rule.

 

If I understand, you wish to minimize this standard deviation. I
believe, the best way is to distribute as many 5 values per bin (from
beginning) until you can distribute exactly 4 values per bins in the
leftover bins.

 

Regards,

Swapnil

 

 

 

++

Swapnil Dwivedi, Sr. Software Engr.

Mentor Graphics, San Jose, CA

Office : +44 (1635) 811-442   (Internal: 8210-1442)

Cell   : +44 (7919) 150-796

Home   : +1  (408)  259-1817 (VOIP)

 

	 

	
________________________________


	From: owner-sv-ec@server.eda.org
[mailto:owner-sv-ec@server.eda.org] On Behalf Of Bresticker, Shalom
	Sent: Sunday, February 11, 2007 4:32 AM
	To: SV_EC List
	Subject: [sv-ec] dividing values among bins

	In coverage bins, where the number of values is not exactly
divisible by the number of bins, the LRM says that the last bin gets all
the extra values. For example, 13 values in 4 bins gets split up as
3,3,3,4. 

	This can create some significant distortions. Suppose you have
499 values in 100 bins. Then they get split up as 4,4,4,4,...,103.

	Wouldn't it have been better to define that they get split up as
5,5,5,5...,4? That way, there would never be a difference greater than
one in the number of values in each bin.

	Shalom

	Shalom Bresticker

	Intel Jerusalem LAD DA

	+972 2 589-6852

	+972 54 721-1033 

	 

	
	-- 
	This message has been scanned for viruses and 
	dangerous content by MailScanner <http://www.mailscanner.info/>
, and is 
	believed to be clean. 


-- 
This message has been scanned for viruses and 
dangerous content by MailScanner <http://www.mailscanner.info/> , and is

believed to be clean. 

-- 
This message has been scanned for viruses and
dangerous content by MailScanner, and is
believed to be clean.