home *** CD-ROM | disk | FTP | other *** search
- Newsgroups: sci.math
- Path: sparky!uunet!mcsun!Germany.EU.net!ira.uka.de!rz.uni-karlsruhe.de!stepsun.uni-kl.de!uklirb!posthorn!vier!neuhaus
- From: neuhaus@vier.informatik.uni-kl.de (Stephan Neuhaus (HiWi Mattern))
- Subject: Serial correlation of uniformly distributed integers
- Message-ID: <neuhaus.726856771@vier>
- Summary: What is the expected serial correlation of uniformly distributed integers?
- Keywords: statistics,serial correlation
- Sender: news@posthorn.informatik.uni-kl.de (News system account)
- Nntp-Posting-Host: vier.informatik.uni-kl.de
- Organization: University of Kaiserslautern, Germany
- Date: Tue, 12 Jan 1993 16:39:31 GMT
- Lines: 46
-
- Hello.
-
- I am interested in the expected serial correlation coefficient of a
- large number of uniformly distributed integers. In particular, let
- N = 2n be a large number, let U_i = Y_{2i} and V_i = Y_{2i+1} for
- uniformly distributed random integers 0 <= Y_i <= M (the precise
- value of M should be immaterial, let's say it is on the order of
- 10^{10}). Compute the serial correlation coefficient C as follows:
-
- n \sum (U_i V_j) - (\sum U_i)(\sum V_j)
- C = ---------------------------------------------------------------
- -----------------------------------------------------------
- / 2 2 2 2
- /(n \sum (U_i ) - (\sum U_i) ) (n \sum (V_j ) - (\sum V_j) )
- -v
-
-
- In _The Art of Computer Programming_, Vol. II, Chapter 3, Knuth writes
- that even with truly random numbers, the correlation coefficient could
- not expected to be exactly zero (this I understand). He then states
- that C should lie between \mu_n - 2\sigma_n and \mu_n + 2\sigma_n 95
- percent of the time, where
-
- ----------
- 1 1 / n (n - 3)
- \mu_n = - -----, and \sigma_n = ----- / ----------
- n - 1 n - 1 - v n + 1
-
- He then goes on: "[These equations] are only conjectured at this time"
- for uniformly distributed random variables.
-
- My question is thus: Have these values been proved by now, if so, are
- there references to the literature, and if not, were these values
- subsequently altered?
-
- Please reply by email, since I don't regularly read this newsgroup.
- If you reply, please send mail to neuhaus@informatik.uni-kl.de, not
- neuhaus@vier.informatik.uni-kl.de. This is a bug in our news
- software, and I'm not sure when we'll get it fixed. Thank you.
-
- Thanks in advance.
-
- --
- Stephan <neuhaus@informatik.uni-kl.de>
- sig closed for inventory. Please leave your pickaxe outside.
- PGP 2.1 public key available on request. Note the expiration date.
-