home *** CD-ROM | disk | FTP | other *** search
- Newsgroups: sci.physics.fusion
- Path: sparky!uunet!zaphod.mps.ohio-state.edu!magnus.acs.ohio-state.edu!usenet.ins.cwru.edu!agate!linus!linus.mitre.org!linus!mbunix!eachus
- From: eachus@Dr_No.mitre.org (Robert I. Eachus)
- Subject: Re: A Little More Data
- In-Reply-To: ames!FNALD.FNAL.GOV!DROEGE's message of Fri, 24 Jul 1992 18:41:53 GMT
- Message-ID: <EACHUS.92Jul24183639@Dr_No.mitre.org>
- Sender: news@linus.mitre.org (News Service)
- Nntp-Posting-Host: dr_no.mitre.org
- Organization: The Mitre Corp., Bedford, MA.
- References: <920724131231.208006aa@FNALD.FNAL.GOV>
- Date: Fri, 24 Jul 1992 23:36:39 GMT
- Lines: 51
-
- In article <920724131231.208006aa@FNALD.FNAL.GOV> ames!FNALD.FNAL.GOV!DROEGE writes:
-
- Pt Gamma Duration Duration Accumulated Estimated Average
- # Status Seconds Hours Net Joules Error-J Power - watts
-
- 1 On 17760 4.9 532.8 49 0.0300 +/- 0.0027
- 2 Off 48060 13.4 1268.8 50 0.0264 +/- 0.0010
- 3 On 87300 24.3 2715.0 100 0.0311 +/- 0.0011
- 4 Off 77220 21.5 1621.6 50 0.0210 +/- 0.0006
- 5 On 82740 23.0 3607.5 50 0.0436 +/- 0.0006
- 6 Off 79980 22.2 2709.7 50 0.0339 +/- 0.0010
- 7 On 83640 23.2 3284.3 50 0.0393 +/- 0.0010
- 8 Off 57300 15.9 1201.0 50 0.0210 +/- 0.0010
-
- Many thanks to Robert Eachus for his data analysis. [Danke.] How
- about it skeptics, what would be convincing?...
-
- I realized last night that a simple binomial test would be fairly
- powerful with this data. The expected result is that the power will
- be greater when the power is on than off. Consider a trial to be
- adding or removing the source, and all we need to look at is whether
- the difference between successive power levels is in the "right"
- direction. So the above data gives us 7 pairs all in the right
- direction. Alpha = 1/2**7 = 1/128 = 0.0078 = 0.78%, so with a one
- percent significance level reject the null hypothesis. (This only
- says that the addition of the radiation source has an effect and it is
- in the right direction, but that is saying a lot. This is already more
- convincing than the statistics in most published papers in many
- fields, including Cold Fusion.)
-
- For completeness the two statisics I posted previously are now:
-
- Using Wilcoxon's W statistic (H0 = no effect, H1 = excess heat during
- on periods): Alpha = 0.057 = 5.7%
-
- Using the signed rank test (this is the less likely to be affected by
- long term drift in the calibration as is the binomial test above):
- Alpha (confidence level) = 0.0625 = 6.25%
-
- Neither is decisive, but I still need more data (for this sort of test
- about 20 data points is usual). I'll do some more analysis over the
- weekend, but at this point unless the data change radically, I'm
- really looking for more information in the data, not whether or not
- there is a treatment effect.
- --
-
- Robert I. Eachus
-
- with STANDARD_DISCLAIMER;
- use STANDARD_DISCLAIMER;
- function MESSAGE (TEXT: in CLEVER_IDEAS) return BETTER_IDEAS is...
-