home *** CD-ROM | disk | FTP | other *** search
- Newsgroups: sci.math.stat
- Path: sparky!uunet!cs.utexas.edu!wupost!darwin.sura.net!mips!news.cs.indiana.edu!orstnews!usenet
- From: choup@bionette.cgrb.orst.edu (Ping Chou)
- Subject: Answers to 'quiz of the day'
- Message-ID: <1992Aug20.044442.6667@talon.ucs.orst.edu>
- Sender: usenet@talon.ucs.orst.edu (Usenet News admin)
- Nntp-Posting-Host: cgrb.orst.edu
- Organization: Biological Computing Consortium, OSU, Corvallis, OR.
- Distribution: usa
- Date: Thu, 20 Aug 1992 04:44:42 GMT
- Lines: 74
-
- The 'quiz of the day' got answered on the same and the following days. Many
- thanks to those of you who responded (so quickly). As for those who asked for
- a summary (and beyond), here it is:
-
- (The quiz, rephrased)
-
- Let P and Q are independent variates of Normal distribution of mean=0 and
- standard deviation= Sp and Sq, respectively. Let R=P/Q.
- What is the distribution of R?
-
- (The answer)
-
- R will follow Cauchy distribution. The normalized density function is
-
- f(R) = (Sr/_pi_)/(R^2+Sr^2),
-
- where Sr=Sp/Sq and _pi_= 3.1415926535897932..., and
-
- INTEGRATE(from a to b)[f(R) dR] = (1/_pi_)*arctan(R/Sr)|(from a to b).
-
- (The note)
-
- 1. Cauchy distribution is the same as Lorentzian distribution. Or may be not,
- but I have done some NMR (sorry, it's Nuclear Magnetic Resonance) spectra
- analysis, by using Lorentzian band to decompose NMR spectra. Both
- distribution have the same density function.
- 2. The parameter Sr in f(R) is analog to standard deviation in Normal
- distribution. Sr is the half width at half height of the peak.
- 3. Let P'=P/Sp, Q'=Q/Sq (which imply Sp'=Sq'=1) and R'=P'/Q', then Sr'=1.
-
- (The most-cited reference)
-
- Advanced Theory of Statistics, originally by Sir Maurice Kendall
- 5th ed., Vol. I, 'Distribution Theory', by Alan Stuart and J. Keith Ord
- (Vol. II also contains many useful stuffs)
-
- (The beyond)
-
- The following table contains accumulated probability of R in 8 intervals,
- (-inf,-10), (-10,-1), (-1,-0.1), (-0.1,0), (0,0.1), (0.1,1), (1,10), (10,+inf)
- from three simulations. Data in each column of 'SIMULATE' are obtained from
- 10^7 R's, and those of 'INTEGRATE' are calculated according to the integral
- expression above.
-
- Accumulated probability of Cauchy distribution of R=P/Q, Sr=Sp/Sq
- ==============================================================================
- Sp=1, Sq=2, Sr=0.5 Sp=1, Sq=1, Sr=1 Sp=2, Sq=1, Sr=2
-
- SIMULATE INTEGRATE SIMULATE INTEGRATE SIMULATE INTEGRATE
- ==============================================================================
- -inf
- 0.016107 0.015902 0.031650 0.031726 0.062803 0.062833
- -10.0
- 0.131533 0.131681 0.218344 0.218274 0.289752 0.289583
- -1.0
- 0.289553 0.289583 0.218205 0.218274 0.131613 0.131681
- -0.1
- 0.062802 0.062833 0.031769 0.031726 0.015813 0.015902
- 0
- 0.062824 0.062833 0.031660 0.031726 0.016073 0.015902
- +0.1
- 0.289866 0.289583 0.218351 0.218274 0.131529 0.131681
- +1.0
- 0.131609 0.131681 0.218253 0.218274 0.289616 0.289583
- +10.0
- 0.015706 0.015902 0.031768 0.031726 0.062801 0.062833
- +inf
- ==============================================================================
-
- Ping
-
- PS. By the time I'm ready to post this, 9 more responses come in with more
- suggestions, hints, references, answers and requests. You boys and girls
- are really beautiful. Thanks again.
-