> > > On Wed, Sep 17, 2014 at 10:45 PM, Jed Rothwell <jedrothw...@gmail.com> > wrote: > >> >> > > >> Miles knew the history of the samples, but he did not tell the groups >> operating the mass spectrometers. >> >> Miles described this in his papers, and I described it in my review: >> >> http://lenr-canr.org/acrobat/RothwellJintroducti.pdf >> >> I suggest you review the review. >> > > >
I read it and was impressed until I became confused by the statistical analysis discussion on page eight. On the one hand you said the duds did not produce any sign of helium above ground: *"Electrolysis power levels during dud runs were raised and loweredconsiderably, from a half-watt to two watts, with no measurable effect on the heliumbackground",* but then you quote Miles, " *For our 33 experiments involving heat and helium measurements, excess heat was measuredin 21 cases and excess helium was observed in 18 studies. Thus 12 experiments yielded noexcess heat and 15 measurements gave no excess helium."* You pointed out in an earlier post that he did not know before hand which were dud cells and which active cells so that means the 33 trials include both dud cells and active cells. If so this implies that some dud cells did produce excess helium since he does not say the 18 cases which produced helium were only associated with the 21 cells which produced excess heat. The validity of his subsequent statistical argument that these numbers constitute evidence of a strong statistical correlation of heat to helium is impossible to gauge unless we are told how many of the dud cells showed helium: *"If one uses these experimental results as random probabilities of P(heat) = 21/33 for excess heat and P(He) = 18/33 for excess **helium, then the probability of random agreement (Pa) for our heat and * *helium measurements * *would be . . . 0.512, and the probability of random disagreement (Pd) would be . . . 0.488. **The presence or absence of excess heat was always recorded prior to the helium * *measurement and was not communicated to the helium laboratory. [The ìblind testî * *procedure.] Based on our experimental results, the random probability of the helium * *measurement correlating with the calorimetric measurement is not exactly one-half. This is * *analogous to flipping a weighted coin where heads are more probable than tails. The * *probability of exactly three mismatches in 33 experiments, therefore, would be . . . 1.203 × * *10^-6 . . . * *The total probability of three or less mismatches in 33 studies would be . . . 1/750,000 . . . * *Furthermore, it is very unlikely that random errors would consistently yield helium-4 * *production rates in the appropriate range of 1011 - 1012 atoms/s per watt of excess power . . . "* Harry