Frye Hearing, June 17, 2013

[cboldt]

Casias v. Florida, 94 So3d 611 (Fla. 2DCA 2011)

FN 2.  We note that at least one statistical methodology--the product rule--has already been found to satisfy the requirements of Frye. See, e.g., Butler v. State, 842 So.2d 817, 829 (Fla.2003); Brim, 695 So.2d at 272.  If Johnson testifies that she used the product rule to generate her statistics, or any other statistical methodology that has already been found to be generally accepted within the scientific community, the State need not prove anew that the chosen methodology satisfies Frye. However, if Johnson testifies to using a statistical methodology that has not been previously determined to satisfy the requirements of Frye, the court will need to hold a full Frye hearing in order to make the necessary determination.

That, in combination with the fact that the product rule is radically incorrect in some cases, forms Mantei's legal argument.  Even if his experts are out to lunch, they pass the Frye test and should be admitted.  What a way to run a trial.

[RickyJim]

I don't really follow what the product rule controversy is all about.  The product rule is correct if the variables or events are independent.  IIRC, in People vs. Collins one of the reasons that the conviction was thrown out by the appeals court is that the state didn't show that independence.  Was that the issue in Casias?

[cboldt]

I don't really follow what the product rule controversy is all about.  The product rule is correct if the variables or events are independent.  IIRC, in People vs. Collins one of the reasons that the conviction was thrown out by the appeals court is that the state didn't show that independence.  Was that the issue in Casias?

No.  I linked the case, you can read what the issue was.  The state didn't lay foundation for how the statistical conclusion was arrived at, i.e., there is no testimony that the product rule, or some other methodology, was used.

FN2 in Casias implies that there will not be an issue of admissibility if the product rule is used, period.  That there is no need to demonstrate or show evidence the accumulated probabilities/variables are independent.

I'm not positive where Mantei is going, unless it is simply to show that the product rule is allowed in Fl, but has been scientifically discredited unless properly qualified (variable independence).  So, if his expert is using something that is scientifically discredited, the testimony can still come in if the method can be valid in some applications.

[nomatter_nevermind]

I'm thinking about dogs that aren't barking. It seems clear to me that the prosecution has no interest in the the 'wind noise' issue, or the moving lights in the clubhouse videos. If they even had any experts look into those we probably would have heard by now, and certainly if they intended to use any at trial.

ETA: I missed the entire hearing. When I came on line an hour or so ago, I checked some of the feeds and and their coverage was done.

[unitron]

...Reich is wrong - the best evidence has 8 bit resolution...

By which you mean, I take it, that 8 bit resolution is the upper limit of the quality of the evidence available in the Zimmerman case, and not that it is superior to (non-existant in this case) 16 bit resolution?

[Cylinder]

By which you mean, I take it, that 8 bit resolution is the upper limit of the quality of the evidence available in the Zimmerman case, and not that it is superior to (non-existant in this case) 16 bit resolution?

Right. Not speaking for cboldt, but the best evidence is that the recording has 8-bit resolution even though the metadata might characterize it as 16-bit. Dr. Wayman explained that is very unlikely that a cell phone recording would have the actual 16-bit resolution that Reich considered so vital to his analysis.

The number of bits represent the resolution of the waveform's amplitude. In an 8-bit recording, there would be 2⁸ - or 256 possible steps in amplitude, whereas a 16-bit would contain 2¹⁶ - 65,536. Dr. Doddington pointed out that the sample used by Reich had half of his asserted 16-bit resolution zeroed. Essentially, someone took an 8-bit sample and recorded it as 16-bits.

[unitron]

Right. Not speaking for cboldt, but the best evidence is that the recording has 8-bit resolution even though the metadata might characterize it as 16-bit. Dr. Wayman explained that is very unlikely that a cell phone recording would have the actual 16-bit resolution that Reich considered so vital to his analysis.

The number of bits represent the resolution of the waveform's amplitude. In an 8-bit recording, there would be 2⁸ - or 256 possible steps in amplitude, whereas a 16-bit would contain 2¹⁶ - 65,536. Dr. Doddington pointed out that the sample used by Reich had half of his asserted 16-bit resolution zeroed. Essentially, someone took an 8-bit sample and recorded it as 16-bits.

It was just that when he said "...the best evidence has 8 bit resolution...", that left a little vagueness as to whether 8 bit was superior to everything else anywhere, anytime, i.e., "8 bit is the gold standard", or whether there wasn't anything available in the Zimmerman case of greater "quality".

(2 bits, 4 bits, 8 bits, no more headroom)

I suppose it could be worse--those 8 extra zeros could have been tacked on as the least significant bits.

(maybe we could call it NRZ encryption)

: - )

[Cylinder]

I think he meant that the best evidence points to the recording being 8-bit - not the 8-bit recording is the best evidence. Reich claims 16-bit. Doddington and Wayman claim 8-bit.

Dr. D pointed out that the metadata characterizes the recording as 16-bit but 8 of those are zeroed. Dr. W points out that voice calls on cellular networks are generally 8-bit signals are and compressed and that a 16-bit recording from a cell call to a 911 operator is very unlikely.

[cboldt]

It was just that when he said "...the best evidence has 8 bit resolution...", that left a little vagueness as to whether 8 bit was superior to everything else anywhere, anytime, i.e., "8 bit is the gold standard", or whether there wasn't anything available in the Zimmerman case of greater "quality".

I meant it is the best that exists in this case.  In a legal sense, best evidence seeks to get closest to the source.  The original is whatever SPD has on their system.

In the digital audio world, some systems use 24 bit data (or more) for greater dynamic range.  Bear in mind that at least Doddington said that 8 bits was sufficient dynamic range.  The nature of speech is that having greater dynamic range (for comparing samples of normal speech) is not needed, nor is a sample rate greater than 8000 Hz needed.

[cboldt]

I thought Wayman's analogy for the looping process was good.  He used a fictitious example of trying to learn what the PIN number was (or account number, or any other number fact), and he increased the amount of data available by repetition of "the first digit is a 1."  No matter how many times you repeat that, you get no closer to knowing what any other digit is.