Continuation Part Eight: Discussion of the Amanda Knox/Raffaele Sollecito case

[codyjuneau]

.
Since the lab stopped using the ABI 7700 after Batch 5, I wonder if it was sent to the manufacturer or one of their representatives for repair and/or calibration? That information should be available from the manufacturer, I would think.

Cody
.

Maybe Stefanoni sent it for cleaning.

.
Seriously though, would that not be irrefutable evidence that there was something wrong with the machine when they used it to test the batch with the bra clasp sample?

Cody
.

 

[Strozzi]

.
Seriously though, would that not be irrefutable evidence that there was something wrong with the machine when they used it to test the batch with the bra clasp sample?

Cody
.

It would be irrefutable evidence that something was wrong, unless the judge does not want to accept it as such. Stefanoni and her department might stonewall you on this, and refuse to acknowledge that the machine was sent for technical examination or servicing. A judge who wants to protect the lab from outside review or scrutiny might decide your questioning is out of order.

 

[Diocletus]

I couldn't pass up doing a little math. Here I have created a little chart showing the calibration curve data for the first 8 runs.[qimg]http://www.internationalskeptics.com/forums/imagehosting/thum_15144539516f392cda.png[/qimg]
(Click on the chart to see the real thing. The thumbnail doesn't show very well)

The formulas that were calculated from the data have the form y= Slope * ln(x) + Intercept. The Slope and Intercept seem to match those calculated by the machine. R² shown in the formulas is a statistical measure of how closely the data fits this curve. These values corresponds to the Fit values computed by the machine.

What is apparent is that there is a natural variation in the Intercept values. The one Diocletus has shown such an interest in (the yellow line at the top) appears to simply be an outlier. But one line shows an apparent abnormality in slope (the blue line for run 566 which crosses the yellow line).

Very interesting, Dan, but I must point out that you missed Run No. 544.

I propose a new experiment. Take all of the RT runs and compare their threshold values. If you happen to see, say 2 outliers on the high side, determine how Stefanoni performed quantification following those outlier runs.

I believe that I have already given you the answer for one of the outliers (Run No. 570): she abandoned usage of the Real Time machine.

After performing said analysis, let's discuss what Stefanoni's actions tell us about what she thought about the "outliers."

 

[Diocletus]

Diocletus, first I want to say that I greatly admire your posts for both their unfailing wit and intelligence. It's always a pleasure to read what you write. But I'm genuinely curious, what are your scientific credentials? We're on the same side of course, so that's not meant as a challenge. I would simply like to know more about what background you possess that positions you for the work you've done on the DNA.

js202, thank you, but allow me answer your question with a question.

What sort of "expertise" are we looking for here? Does it really require an expert to figure out that the quantity for a sample is 5x greater than what it should be? An idiot could figure this out if he knows how to read the printout, and frankly, the printout isn't complicated if you try to understand it. (Apologies to Bill Williams). Same with the negative controls that we previously identified as contaminated: everybody knows that if something isn't supposed to have anything in it, then its a problem if it does have something in it.

How about the transposition of subrep numbers? This just requires one to understand that 6 comes before 7 and "a" comes before "b." A kindergartener could figure this out.

How about the issue of plate numbering? The whole methodology is spelled out, and its not that complicated, either: make a list of the positive quantifications and match them up to ID numbers and egrams. As a workcheck, the whole system corresponds perfectly with the known egrams. This isn't "science" or really even technical. It's just figuring out how pieces of a puzzle fit together based on sequences.

None of this is that complicated. In fact, I would argue that a scientist might be the wrong type of person to figure that out, because scientists like to figure out the hows and the whys, and might accept results that are in the "gray" if they can't figure out why they are wrong. If I were approaching this type of analysis, I might look for someone with the following skills:

1. Able to understand, on a very basic level, the process, and how the pieces of data generated during the process fit together

2. Experienced in forensic examination of data: figuring out where pieces of data are missing

3. Has seen, understands and can detect document production fraud

4. Critical, and not accepting of what someone says about their own work

 

[Bill Williams]

js202, thank you, but allow me answer your question with a question.

What sort of "expertise" are we looking for here? Does it really require an expert to figure out that the quantity for a sample is 5x greater than what it should be? An idiot could figure this out if he knows how to read the printout, and frankly, the printout isn't complicated if you try to understand it. (Apologies to Bill Williams). Same with the negative controls that we previously identified as contaminated: everybody knows that if something isn't supposed to have anything in it, then its a problem if it does have something in it.

How about the transposition of subrep numbers? This just requires one to understand that 6 comes before 7 and "a" comes before "b." A kindergartener could figure this out.

How about the issue of plate numbering? The whole methodology is spelled out, and its not that complicated, either: make a list of the positive quantifications and match them up to ID numbers and egrams. As a workcheck, the whole system corresponds perfectly with the known egrams. This isn't "science" or really even technical. It's just figuring out how pieces of a puzzle fit together based on sequences.

None of this is that complicated. In fact, I would argue that a scientist might be the wrong type of person to figure that out, because scientists like to figure out the hows and the whys, and might accept results that are in the "gray" if they can't figure out why they are wrong. If I were approaching this type of analysis, I might look for someone with the following skills:

1. Able to understand, on a very basic level, the process, and how the pieces of data generated during the process fit together

2. Experienced in forensic examination of data: figuring out where pieces of data are missing

3. Has seen, understands and can detect document production fraud

4. Critical, and not accepting of what someone says about their own work

I'm standing right here, y'know!?

 

[anglolawyer]

js202, thank you, but allow me answer your question with a question.

What sort of "expertise" are we looking for here? Does it really require an expert to figure out that the quantity for a sample is 5x greater than what it should be? An idiot could figure this out if he knows how to read the printout, and frankly, the printout isn't complicated if you try to understand it. (Apologies to Bill Williams). Same with the negative controls that we previously identified as contaminated: everybody knows that if something isn't supposed to have anything in it, then its a problem if it does have something in it.

How about the transposition of subrep numbers? This just requires one to understand that 6 comes before 7 and "a" comes before "b." A kindergartener could figure this out.

How about the issue of plate numbering? The whole methodology is spelled out, and its not that complicated, either: make a list of the positive quantifications and match them up to ID numbers and egrams. As a workcheck, the whole system corresponds perfectly with the known egrams. This isn't "science" or really even technical. It's just figuring out how pieces of a puzzle fit together based on sequences.

None of this is that complicated. In fact, I would argue that a scientist might be the wrong type of person to figure that out, because scientists like to figure out the hows and the whys, and might accept results that are in the "gray" if they can't figure out why they are wrong. If I were approaching this type of analysis, I might look for someone with the following skills:

1. Able to understand, on a very basic level, the process, and how the pieces of data generated during the process fit together

2. Experienced in forensic examination of data: figuring out where pieces of data are missing

3. Has seen, understands and can detect document production fraud

4. Critical, and not accepting of what someone says about their own work

I doubt you would stick your neck out if not sure of your ground and I also know that Tom Zupacnic endorses your conclusions. I also know from your posts you are someone deserving respect. None of which is a substitute for understanding this myself. I am in the idiot's corner with Bill until I do. No big deal. It wouldn't be the first time. Would you please explain:

1 what a 'plate' is?
2 At what point in the DNA profiling operation are plates used?
3 What do they look like - something like the ice cube tray in the freezer?
4 what are these ID numbers?
5 when are they applied to an item and what type of item (a knife, a sock, a sample from a blade etc)
6 describe the process 'at a basic level' and say how the pieces of data fit together

Treat this as a request for further and better particulars. I will apply for an order in case of default

 

[js202]

js202, thank you, but allow me answer your question with a question.

What sort of "expertise" are we looking for here? Does it really require an expert to figure out that the quantity for a sample is 5x greater than what it should be? An idiot could figure this out if he knows how to read the printout, and frankly, the printout isn't complicated if you try to understand it. (Apologies to Bill Williams). Same with the negative controls that we previously identified as contaminated: everybody knows that if something isn't supposed to have anything in it, then its a problem if it does have something in it.

How about the transposition of subrep numbers? This just requires one to understand that 6 comes before 7 and "a" comes before "b." A kindergartener could figure this out.

How about the issue of plate numbering? The whole methodology is spelled out, and its not that complicated, either: make a list of the positive quantifications and match them up to ID numbers and egrams. As a workcheck, the whole system corresponds perfectly with the known egrams. This isn't "science" or really even technical. It's just figuring out how pieces of a puzzle fit together based on sequences.

None of this is that complicated. In fact, I would argue that a scientist might be the wrong type of person to figure that out, because scientists like to figure out the hows and the whys, and might accept results that are in the "gray" if they can't figure out why they are wrong. If I were approaching this type of analysis, I might look for someone with the following skills:

1. Able to understand, on a very basic level, the process, and how the pieces of data generated during the process fit together

2. Experienced in forensic examination of data: figuring out where pieces of data are missing

3. Has seen, understands and can detect document production fraud

4. Critical, and not accepting of what someone says about their own work

Thanks for your thorough and thoughtful reply. It sounds as though you consider what you did an exercise in process and logic, rather than science, per se.

What would you think about the creation of a kind of map key for your page?

 

[Diocletus]

1 what a 'plate' is?

Generally, a "plate" is a set of sample sequences. They contain 16 samples, or 16 plus some multiple of 4, up to 32 (highest observed):

Dr. Stefanoni answered that firstly, there were two sequencers in the laboratory, one with four capillaries and one with sixteen capillaries, both used interchangeably; she added that, theoretically, it is not possible to eliminate the risk of contamination. She did, however, assert that this risk is mitigated via the observance of proper laboratory procedure, and subjecting all the instrumentation and material to regularly programmed technical check-ups.

Massei pp. 218-19

2 At what point in the DNA profiling operation are plates used?

After successful quantification.

3 What do they look like - something like the ice cube tray in the freezer?

They can, yes.

4 what are these ID numbers?

A sequential set of numbers that identify the amplifications within a plate.

5 when are they applied to an item and what type of item (a knife, a sock, a sample from a blade etc)

They are applied at amplification, and applied to every sample that successfully quantified.

6 describe the process 'at a basic level' and say how the pieces of data fit together

http://murderofmeredithkercher.com/missing-profiles-draft/

 

[Desert Fox]

I'm standing right here, y'know!?

I think I understand the arguments but the arguments need some kind of "punch out" to get a jury or judge to understand it.

 

[Diocletus]

What would you think about the creation of a kind of map key for your page?

Yes, that is definitely needed. The whole exercise became much more involved than originally expected . . . due to consistent discovery of lab problems in each succeeding batch.

 

[js202]

Yes, that is definitely needed. The whole exercise became much more involved than originally expected . . . due to consistent discovery of lab problems in each succeeding batch.

I think it would be important to making the case you've put together read as unassailable. I, personally, am just now beginning to penetrate meaning and put together a coherent narrative based on data that initially read as hieroglyphics (despite my time constraints, I have circled back to your document two or three times to make this effort).

And, btw, just so it doesn't get lost in all the niggling, kudos and congratulations on the work.

 

[Dan O.]

Very interesting, Dan, but I must point out that you missed Run No. 544.

I didn't miss it. I read those hand printed digits as 566. I had first interpreted the number as 544 but later thought it was closer to 566 and changed it.

I propose a new experiment. Take all of the RT runs and compare their threshold values. If you happen to see, say 2 outliers on the high side, determine how Stefanoni performed quantification following those outlier runs.

Sounds like a plan that can lead to confirmation bias. Has there been any literature search to see what values are acceptable?

The most important output in terms of determining that the machine is working properly appears to be Fit. This is a measure of how closely the calibration data fits a straight line defined by the Slope and Intercept. Of the data I've examined, Fit has always been either 0.99 or 1.00 (rounded by the machine). If the machine were malfunctioning to the point of presenting inaccurate results, this Fit should drop precipitously.

Maybe there was a run where Fit fell. I suspect we would never see that result because that would imply that the lab tech running the machine made a mistake and such is not permitted in the perfect world of Italian forensics.

I believe that I have already given you the answer for one of the outliers (Run No. 570): she abandoned usage of the Real Time machine.

After performing said analysis, let's discuss what Stefanoni's actions tell us about what she thought about the "outliers."

You don't even know if that Threshold is outside the normal range. Let's look at everything, consult the literature and consult the professionals before making conclusions.

 

[AmyStrange]

half fast logic at it's best

-

Bill Williams responding to Stilicho:

Stilicho, nobody witnessed a break-in. If they had, people would have come forward the next day. To think that they wouldn't have come forward is to just show contempt for the many good people of Perugia (who have the misfortune to live in a city with incompetent police, a sadistic presecutor, and crony judges).

What they witnessed were two dogs mating or a drunk or drugged man passed out on the ground. I'm glad that Curatolo was able to get to his feet, get up the staircase, resume his customary position on the park bench, and then stay awake reading for two hours to witness Perugia's Crime of the Century!

Doesn't it make you wonder how much Mignini's case rests on testimony from unreliable witnesses. And now Stilicho thinks all these good people saw something and the next day didn't put 2 plus 2 together and come forward. Sheesh . . .

-

No one saw or heard the staged break-in either, so that mustn't of happened either. So neither must have happened... and yet, the window was broken the next day. Go figure,

d

-

 

[Diocletus]

Sounds like a plan that can lead to confirmation bias. Has there been any literature search to see what values are acceptable?

Dan, you didn't play the game. Here goes:

There are two runs where the threshold value exceeded .10:

Run No. 544 (6 Nov. 2007) and Run No. 570 (3 Jan. 2008).

We know that after Run No. 570, Stefanoni abandoned the Real Time machine and thereafter quantified everything on a Fast Time machine.

As to Run No. 544, immediately after running this batch (in fact on the same day), Stefanoni abandoned the Real Time machine and started using a Qubit Flourometer to perform quantifications. She did not resume use of the Real Time machine (for this case) until 26 Nov. 2007, and then used it only until Run No. 570, described above.

So, did Stefanoni believe that her machine had a problem? It's my recollection that she actually testified to that effect . . . can anyone locate the relevant testimony (i.e., why she started using the Qubit)?

 

[Chris_Halkides]

evaluating real time PCR

Table 1. Performance evaluation of real-time PCR.

Factors Recommendations Criteria
Efficiency Serial dilution with 5-log dilutions Slope: ~ –3.3R2 >0.99
Precision Minimum of 3 replicates Standard deviation<0.167
Sensitivity High replicate number of reactions Statistical test analysis

Table is modified slightly due to formatting issues. The high replicate number is necessary "for low copy number sample input due to Poisson distribution"
Link

 

[AmyStrange]

Someone should just file a lawsuit...

-

The jurisdictional issue is a problem for getting huge cash awards back. The more likely outcome would be barring the company from delivering its product into the region where there are insufficient safeguards to insure their will be no harm caused by its improper use.

-

File a suit against the manufacturer for a 100 million, serve notice to the manufacturer, and then file a press release with the media.

The suit may get quashed on summary judgement, but at the least, the press release might get something started.

Plus, in order to quash the suit, the manufacturer has to send someone to the court to ask for a summary judgement, and if no one appears for the defense, isn't the suit automatically won (in most cases) by the plaintiff, by default?

Just my opinion,

d

-

 

[Bill Williams]

I doubt you would stick your neck out if not sure of your ground and I also know that Tom Zupacnic endorses your conclusions. I also know from your posts you are someone deserving respect. None of which is a substitute for understanding this myself. I am in the idiot's corner with Bill until I do. No big deal. It wouldn't be the first time. Would you please explain:

1 what a 'plate' is?
2 At what point in the DNA profiling operation are plates used?
3 What do they look like - something like the ice cube tray in the freezer?
4 what are these ID numbers?
5 when are they applied to an item and what type of item (a knife, a sock, a sample from a blade etc)
6 describe the process 'at a basic level' and say how the pieces of data fit together

Treat this as a request for further and better particulars. I will apply for an order in case of default

Anglo.... this is what separates innocenters from guilters, in my opinion. To echo what you said.....

I have nothing but respect for Chris_Halkides, Tom Zed, and Diocletian. My respect for them comes from me not understanding anything they say.

Yet people who i.d. themselves as experts, seem to agree with them. And what guilters have in reply is ad hominem - and this whole saga judicially has not moved from the fact that Massei believed Stefanoni on her say-so, not because of independent review.

And yet - what separates us from the heavy-browed set, is that if I were on a jury listening to what has transpired in the last few days, my vote would be to discount it.

And my own confirmation bias is to believe it. It would be nice to unquestionably believe it. It certainly makes sense in the "osmotic" view I've adopted.....

..... but I cannot accede to something I cannot understand.

I do understand that women do not have Y-genetic material. They told me that in science class the year I graduated highschool. Apparently no one told Nencini. Otherwise I know nothing of twin helixes - I cannot off the top of my head tell you how many human gene-pairs there are.... although, I seem to remember they come in pairs.

But as it rests I would have to simply set this aside - and the problem is not science, it is me.

I'm desperate not to get drawn into things the way guilters do. Mignini says it, so it must be true. Stefanoni said it, so Massei sends two kids to jail for a quarter century.

I may move to Missouri.

 

[Diocletus]

I didn't miss it. I read those hand printed digits as 566. I had first interpreted the number as 544 but later thought it was closer to 566 and changed it.

So, getting back to this for a moment, the two runs with the problematic thresholds (570 and 544) are both in the "outlier" space, and 544 has the screwed up slope that you noted earlier. Hmm.

 

[anglolawyer]

Generally, a "plate" is a set of sample sequences. They contain 16 samples, or 16 plus some multiple of 4, up to 32 (highest observed):



Massei pp. 218-19



After successful quantification.



They can, yes.



A sequential set of numbers that identify the amplifications within a plate.



They are applied at amplification, and applied to every sample that successfully quantified.



http://murderofmeredithkercher.com/missing-profiles-draft/

Thank you. Proceedings are stayed while I look into this. You may resume your seat counsellor

 

[Dan O.]

So, getting back to this for a moment, the two runs with the problematic thresholds (570 and 544) are both in the "outlier" space, and 544 has the screwed up slope that you noted earlier. Hmm.

But according to Chris, the slope of 544 is the only normal one showing a 100% efficiency (slope = -3.31). All of the others are hyper efficient.