I would also like to add that they should calculate confidence intervals on the active runs so we would know the 95th percentile lower bounds for those runs. The standard deviation for the active runs could be different from the calibration runs.
On Thu, Jun 27, 2013 at 5:44 AM, Jack Cole <jcol...@gmail.com> wrote: > Eric, > > My understanding is that +.25W is the 95th percentile for the EU cells and > +.5W is the 95th percentile for the US cells. They are using two sided > confidence intervals (+/-.25W) so this would be +/- 1.96 standard > deviations at .25W. They don't present the actual SD here: > http://www.quantumheat.org/index.php/en/follow/follow-2/285-us-eu-cell-calibration-results > . > > We can get it by working backwards. (.25/1.96=.13) > > Now we can get the standard deviations equivalent of 2.5W. (2.5/.13=19.2) > > So for the EU cell, they are 19.2 standard deviations above calibration. > I can't find a conversion chart with a fairly quick look that goes above 6 > standard deviations. 6 SD is equivalent to 99.99999975 percentile. By > error and chance alone, this would indeed be a very rare occurrence. > (Expected to occur less than 1 time out of 10 billion runs with an SD > multiple of 6, so use your imagination for 19.2 SD multiples). > > That does not rule out some type of systematic error, which should be the > first assumption. But if systematic error can be ruled out, this is indeed > a robust finding. > > Jack > > > Jack > > > > On Wed, Jun 26, 2013 at 9:28 PM, Eric Walker <eric.wal...@gmail.com>wrote: > >> On Wed, Jun 26, 2013 at 12:05 PM, H Veeder <hveeder...@gmail.com> wrote: >> >> Both the EU Cells and the US Cells were switched on and BOTH indicated >>> excess energy as the cells came to equilibrium at higher temperatures than >>> during the calibration tests. The EU cell with the active wire was >>> indicating up to 2.5W of excess power over the 30.4W input power (~6% >>> excess). That is well above the 95% confidence limits for that cell >>> (~0.25W). The US Cell was indicating approximately 1.4 watts excess, >>> again, well above the ~0.5W confidence interval. Very exciting to see >>> something positive and especially simultaneous. >>> >> >> It is encouraging to hear that MFMP are seeing excess heat. But we >> should not get too excited yet; 2.5 W and 1.4 W are small values, and 95 >> percent confidence is only two standard deviations from noise. I have >> heard that scientists often look for 5 standard deviations (5 sigma). For >> the MFMP calorimeters currently being used, with the glass and the SB >> equation, I suspect it will not be that convincing for people until they >> see 10-20 W excess heat (integrated excess power, including periods of >> endotherm). I recall Paul Hunt saying they needed very convincing results >> with the glass assemblies for them to be convincing to anyone. >> >> The control cells in each location are performing at or below calibration >>> values. >>> >> >> Perhaps someone here knows -- is it a problem if a control cell performs >> below the calibration values? >> >> The internal cell temperatures seem to be slowly degrading, but the >>> external cell temperatures are holding steady. >>> >> >> Another weird thing to try to understand. >> >> Eric >> >> >
