The scientific consulting firm that the NFL hired is comical (the 'pay to get the results you want' firm, i believe the same firm that said 2nd hand smoke doesn't cause negative health effects):
https://www.reddit.com/r/nfl/comments/3fucbk/from_the_brady_appeal_exponent_statisticians_and/
Q. Dr. Marlow, the variability analysis that
23 took into account the post-game data that you wanted
24 to include in the report but didn't get included in
25 the report, your view that wasn't included in the
1 report would have been more prejudicial to the
2 Patriots, correct?
3 A. Yes, yes, and that was part of how I
4 interpreted why I was overruled is we didn't want to
5 be too aggressive in it. We were trying to be fair.
6 Q. In your view, based on the variability and
7 analysis, taking into account the post-game data
8 would have made it more likely that the Patriots'
9 balls did not start at the same psi level when they
10 were introduced on the field and after Walt Anderson
11 had gauged them, correct?
12 A. Yes. The very -- the most natural
13 explanation for the variability is that they started
14 at different pressures, yes.
15 Q. And is it fair to say that this was not
16 included in the report because there was a consensus
17 between you and Exponent that it was better to take
18 a conservative approach and not use the post-game
19 data because it might open you up to some criticism,
20 much of which we have heard today?
21 A. Yes, that's fair. That's why I went along,
22 but I couldn't contain myself here; I'm sorry.
23 MR. REISNER: Nothing further.
__________________
5 Q. Now, let me ask you this. You mentioned that
6 you tried to replicate Dr. Snyder's work, correct?
7 You believed you were able to do that?
8 A. I believe I was able to do that.
9 Q. And you pointed out that by doing some
10 variability analysis for the second measure, you
11 could achieve a statistical significance that's
12 above ten percent statistical level, not a five
13 percent, right?
14 A. The p-values I calculated were between five
15 and ten percent.
16 Q. It would not be significant at the
17 five-percent level, right?
_______________
7 Q. I will ask you this: Have you ever done a
8 case you can recall where you had four observations
9 was your entire data set and that's all, or less?
10 And if you can't identify, "no."
11 A. I have reviewed evidence that has been put
12 forth where people have taken one or two
13 measurements and tried to reason on that basis.
14 Q. And you think that's a proper thing for a
15 statistician to do with one or two observations?
16 A. Typically, it is not. Typically it is not
17 because there is uncertainty. But if you had a
18 situation where there was no variability, you had a
19 population that had no variability in it and you
20 wanted to learn about that population, one
21 measurement would be enough because there is no
22 variability.
23 MR. KESSLER: I don't have any further
24 questions. You can keep going if you want, but I'm done
___________________
11 Q. Looking just at the halftime data which you
12 did include, there is no statistically significant
13 effect, correct?
14 A. That's correct. 
Facebook
Twitter