Yesterday, you were kind enough to submit to a deposition at Mr. Baker's law office, with being present -- me being present, Mr. Lambert being present; is that right?
And at that deposition, you explained that the assumptions that you used to come up with the figures that you've presented were assumptions chosen by yourself correct?
And in doing these calculations, you did a whole different -- all sets of possible combinations of racial groups, to come up with a range from one number to another number in terms of how common these patterns might be. Is that a fair way to describe that?
And what you tried to do was to use that combination that would produce the number that is the most frequent number. In other words, the number that would be most helpful to Mr. Simpson, for instance, on one end of the range, and on the other end of the range, would be the more extreme number that would be more incriminating; is that correct?
I think I understand.
I did a series of California calculations according to different racial backgrounds of the unknown contributors and chose the ranges, the smallest and the largest, yes.
What you were trying to do is come up with the figures that would be most favorable to the defense and least favorable, correct?
No. I'm just simply giving the range of numbers. I could have presented the whole 16 numbers.
The bottom end of the range is the number that is the most common appearance of that particular pattern, correct?
Now, you have particular expertise in calculating statistics for mixtures of different DNA patterns, correct?
Is it accurate that you consider yourself to be perhaps the world's foremost expert on that topic?
Is there anybody you would put above you, in terms of level of knowledge and ability, to calculate these kinds of numbers?
I don't think I want to play that kind of a game. But there are several people who do these calculations just as well as I do.
KEY QUOTEA Dr. Evett in the United Kingdom, Dr. Buckleton in New Zealand, I would say, are at least as good as I am.
I don't think I should answer that because the -- the situation is conceivable. Anybody who thought about this issue, who's a competent statistician, would be at least as good as I am. So to put me on a pedestal merely because I've thought about it is a little unfair.
Okay.
Is it fair to say that with DNA testing, it's broken down into three basic steps: The first one being the generation of an Autorad or a testing strip; the second one being the interpretation of that to determine whether it matches the evidence -- matches the suspect; and then the third step being what you do is calculate the frequencies of the possibility that it might be someone else, as opposed to the defendant, that resulted in that match; is that correct?
No. I have to restate -- state the last statement. These numbers don't talk about the probability of coming from someone else; these frequencies talk about the frequency or the probability of seeing this probability, if it comes from someone else.
It's very easy to get confused in the terminology that's used in this, with respect to mixtures, is it not?
Would you agree with me, the three steps that I described, it's a logical way to break down the analysis of DNA testing results?
And you have nothing to do with the first part; that is, nothing to do with what goes into generating an Autorad, collecting evidence, or generating a DQ alpha testing, for instance?
And in your analysis, you don't have anything to do, really, with the second step of determining whether or not there is a match to a defendant in a case; is that correct?
Okay.
So your area of expertise is the third step; that is, calculating these frequencies that you've talked about?
Now, where you're talking about an evidentiary stain that only has two alleles, that's an indication that the source of that is one person, correct?
And calculating a frequency for that situation is a lot easier than calculating frequency for mixture; would you agree with that?
I -- to a -- only to a degree. I think the concepts are the same. We ask the question; here's the profile. What's the chance of that it has this type, if it came from an unknown person? That's the single stain.
For mixed stain, we say here's the profile. What's the chance it has this stain? It came from two unknown people or three.
So it's really the same question.
But with the single stain you're talking about, what's the chance that if I go out into the population of Caucasians, for instance, and pick somebody at random, that they would have that same pattern -- correct?
Now, where you're talking about mixed stains where there may be two, three or four contributors, it gets more complex in terms of the number of possible alternatives that you could have, that could account for those results, other than the defendant contributing the sample, correct?
Well, it's still the same question.
We just have more -- more people that could have contributed, so I don't see a logical distinction.
But you're certainly right; there are more combinations in the sense of different combinations of the different racial groups.
And I say "racial group" as the shorthand for the different data bases I've been using.
And if we take an example of a DQ alpha result that has four alleles, this -- let's say the 1.1, the 1.2, the 1.3 and the 4 -- tell me how you would analyze the number of -- the possible numbers of people that could have created that stain.
And so let's start out with that.
We know that couldn't have come from one person -- one person.
So the next step would be to say, what's the chance of seeing these four alleles if they came from two people. And then we need to say what alleles do these two people have? And that's alleles 1, 2, 3, and 4 for the moment. What's the chance that two people have these four alleles?
Well, the first person could have 1 and 2, and the second person could have 3 and 4; that would work. Or the first person could have 1 and 3 and the second person could have 2 and 4; that would work. And we just go through a list.
And I think there's about six different combinations of the way that two people between them would have this particular set of those four alleles.
Now, if you wanted to take into account the number of -- possible combinations of three people that could create that would be consistent with those four alleles, how would you go about doing that?
Same question, same procedure.
We say here's four alleles, 1, 2, 3, 4. And I want to ask the question, how likely is it that I see those four when I take three people. So I take three people and I say, what does the first one have.
Well, the first one could have 1 and 2; the second person might have a 3 and a 4; and then the third person doesn't have to do anything else for the mixture because I've already got the four alleles accounted for; so this third person could have any combination of those four alleles.
The important thing is that person could not have any additional allele. DQ alpha has six alleles, so that third person couldn't have a 5 and a 6.
So, once again, it's not difficult to say in words; it's a little time-consuming to write down all the combinations.
And that's what I do. That's what I used to do a year ago. And that is error prone when you write down a lot of combinations, it's easy to leave some out. So with a little careful thinking, we came up with a formula which said if these are the alleles, we need to account for -- if these are the number of people that we think account for them, here's an equation, a formula which will tell you the probability.
Now, you can do the same kind of analysis if there were four people that contributed to that stain that has four alleles?
And then what you do, in order to come up with the range of numbers that you've presented here, is to take each one of those possible alternatives, and take all possible combinations of racial groups that might fit into that combination of two people, three people or four people, and calculate how common or what's the frequency of that particular pattern in the population, and in the particular population you choose, right?
And you come up with a whole series of numbers; if you've got 16 possible combinations, you could have 16 different numbers, couldn't you?
Well, as I say, that -- yes, but that's not the way we now do it, we don't have a long list of numbers, we have an equation which spits them out for us.
You take the ends of it, you take the one that's most favorable to the defense; in other words, the one that's the most common possibility and then the one that's the least common possibility, or the most rare?
That's not the first table that you generated in preparation for your testimony today, is it?
Let me show this to you first, and tell me if this is the one, the table that you generated before you generated this one?
The table that you presented here today has the upper range number as 1 in 300 billion, correct?
Yes. The one extremely large number is 50 percent bigger than the other extremely large number.
The difference between the 300 billion is simply due to the difference that you rounded or truncated numbers, right?
It's the same underlying data, but it gives you different results to the extreme of 100 billion possibilities, correct?
When I generated the first number, I truncated the numbers then I did when I generated the second number, it's the same underlying data.
The numbers were different because you dropped some digits off on the first one that you didn't on the second, correct?
No, no. The numbers that I put into my program to calculate the formula were the same, and the numbers that came out were the same numbers. I copied down off the screen, on my yellow piece of paper to do the multiplication when I truncated and when I didn't, and that was probably a foolish thing to do, trying to be overly conservative.
You wouldn't have seen this difference had I rounded -- I took numbers like a 1.9 and called it a 1 on Tuesday and yesterday I called it a 1.9. Before I did the multiplication, if I had rounded I would have called it a 2, and we wouldn't have seen this difference.
I was being -- I guess I was being lazy on Tuesday.
Did Mr. Lambert or anybody tell you that you shouldn't be quite so conservative and ask you to generate another table that was a little bit more --
How much time did you put into preparing -- doing your calculations and preparations for your testimony in the criminal trial?
And about how much time have you put into the calculations that you've done for the plaintiffs in this case?
Well, the immediate calculations have been just over the last week, but I've had a year to write out the theory, to think about things, so that these numbers reflect much greater theory than gestation.
(BY MR. BLASIER) Let me give you a copy of this.
I'd like you to take a look at that document and compare it to your table.
And would you agree that in the left-hand column under "Item," we have the same sets of stains that you have in your table, and just in slightly different order?
And what we're talking about here, so it's a little clearer, G10 is a stain from the Rockingham glove, correct?
And now you have testified in this case -- in the column that you've labeled "Civil," this is the lower range numbers for these various stains that you've testified to; is that correct?
The numbers up there, this is what you -- the hypothesis that you adopted, the assumptions that you made to come up with numbers that you felt were most favorable to the defense, correct?
The numbers that I have here in the "Civil" column are the frequencies that you've presented here that are the lower ends of the range that you felt, in doing your calculations, were the ones that were most favorable to the defense?
Dr. Weir, do you remember testifying in your deposition when I asked you -- (Reading:) Q. Why did you only do calculations for pairs of people rather than for also three contributors and four contributors?
Your answer was: Because this is the more conservative number.
Remember saying that yesterday?
Yes, I said that, and as I said in my direct examination, that's generally true. There are exceptions to that, particularly for the RFLP where there are some unseen bands.
No, I didn't, but as a result of you asking me that, I thought about it again and I actually did those calculations, and we can go through them if you wish.
Well, I think we will.
And you said also in the deposition that, when questioned -- (Reading:) Q. So when you say more conservative?
Your answer was -- (Reading:) More favorable to the defense.
Correct?
Now, what you're telling us now is that these numbers aren't necessarily the ones that are the most favorable to the defense, correct?
No. What I said on my direct examination, and again on cross, is that these numbers are the most conservative assuming two contributors. That's my testimony.
And you didn't do those calculations until I questioned you -- until after I questioned you yesterday, correct?
You were trying to come up with the number, the lower range that were most favorable to the defense?
That's right. And as a general rule, that's what happens. I would be happy to make other calculations.
You're the one who chooses which assumptions to make which calculations to present to this jury?
Yes, I did the simplest ones consistent with the evidence, which has 4 alleles to the probe, which suggests two contributors.
Now, in the criminal trial, you testified, did you not, that the frequency most favorable to the defense for G10 for the PCR results was 1 in 3900; isn't that correct?
Let me refer you to 33775 of the criminal trial transcript, starting at line 27, would you look at that to yourself, you can look above it too, as well, and tell me if you did not, in your direct testimony in the criminal case, state the frequency of G10, the most favorable to the defense, as 1 in 3900?
Well, it's not laid out here. Looks like the 3900 goes -- yes, it does say there are two contributors, that's right.
Let's write 3900 under G10 criminal trial number one.
Now, during the course of your cross-examination in the criminal trial, the defense made you aware of the fact that you had made a mistake, hadn't you?
And what you did was, you didn't count all the possible combinations that might result from the presence or absence of the 1.2 allele in the DQ Alpha system; isn't that correct?
That was as a result of a fundamental misunderstanding that you had about the DQ Alpha system and the 1.2 allele, isn't it?
No, I think the record is clear that that's exactly wrong. I have a complete understanding of that at this time. When I read the program in the hotel room, I left out a line.
So you understood at the time that where there was a 1.2 allele in the DQ Alpha system, that sometimes you could tell it was there and sometimes you can't; isn't that correct?
So you went back and recalculated the numbers after the defense brought to your attention your mistake and presented a second number, 1 in 1600, isn't that correct, for G10?
Let's cross out the 3900 because that's an error and let's put 1 in 1600 which is the second number you gave in the criminal trial, okay.
And the reason for that error was you didn't count the 1.2 allele combinations that could have been there. Correct?
Let's put 1.2000 (sic) there, shorter band. That's the reason for that error.
Now, Dr. Weir, your testimony today in the civil case is that that frequency that's most favorable to the defense is 1 in 2600, correct?
And that's because this number that you testified to under oath in your cross-examination in the criminal trial was wrong, correct?
Yes. And I was -- I think we talked some time to explain the difference between appropriate numbers and wrong numbers. The number is correct. And I'm sure you recall that I said it was not a mistake.
I asked you a question, is it a proper way to do it, to count that combination twice, which is the reason why -- for the difference of 1 in 1600 and 1 in 2600, and you said no. Correct?
That's not the proper way to present it. The calculations were correct. I think it's better to present it the way I have now.
You counted one set of alternatives twice, that's why this number is almost half of this number, isn't that true?
The calculations are correct. I don't think it was the proper way to present it. I notice, however, that these numbers are consistent, they both fall in the ranges of the others I've already presented.
So let's put down, up here, reason, in red, for that improper number.
Now, in the criminal trial, on your direct testimony, for stains 303, 304 and 305 PCR, you testified that the frequency most favorable to the defense in that case was 1 in 1400, didn't you?
Why don't you write 1400 -- 1 in 1400 in that column.
Then, after you were cross-examined in the criminal trial, you recognized that you had made the same mistake with respect to the 1.2 allele on that number, isn't that true?
Dr. Weir, you recognized that you made the same mistake with that number that you had made with G10, correct?
I changed my computer program once, one line, one program, which changed all the DQ Alpha results.
All the numbers involved in DQ Alpha were wrong for the same one line in one computer program.
Now, when you calculated again for your cross-examination in the criminal trial, you presented a number of 1 in 570 as the now correct number, isn't that accurate?
Let's put 1 in 570. Then this was the 1.2 error. Let's write 1.2 over here. Let's cross out this number and let's cross out this number too.
Now, you've testified that the proper number for 303, 304, 305 for the PCR results is 1 in 960, correct?
That's because of the double issue, counting something twice that shouldn't have been counted twice, correct?
Isn't it accurate, that for stain 303, 304 and 305, the number most favorable to the defense was not any of these numbers, but was the number that you derived if there were three possible -- I'm sorry -- four possible contributors; isn't that correct?
No, because that number is not the appropriate one at that time -- at that time it was, yes.
Okay.
Let's write 1 in 41 here. And let's just write MC, for multiple contributors.
MR. P. BAKER: Under the reason?
I wrote MC for multiple contributors.
We get a more favorable number from you where more than --
It's not a number I'm testifying to here. It's not the appropriate number for mixed contributors.
That's not -- that was the correct number that had the double issue. I would no longer talk about it now.
But that's not a number -- you're misstating my -- my testimony here this morning. I haven't testified to a four contributor number here.
The number that's most favorable to the defense is the situation where you have four contributors, as you testified at the criminal trial, and that number, at that time, was 1 in 41, which is now wrong, correct?
At the criminal trial, you testified that the most frequent number -- the most frequent situation where you would get that pattern from picking people from the population at random was with four contributors and you would get that pattern 1 in 41 times, wouldn't you?
So it's now 1 in 6400 for four contributors.
What's the difference, the double issue, the error you made in the criminal trial?
For G11 and G13, two glove stains, the PCR number you testified to in the criminal trial, I'll represent to you, was 1 in 14. Sound right?
Two, I believe.
1 in 14 here, please.
And again you made the same error on that one that you made on the one above, it is double -- I'm sorry, the 1.2 error, I'm sorry, you did not make that error on that one. That doesn't have 1.2 allele.
You're not testifying that the number most favorable to the defense is the 1 in 27 rather than the 1 in 14, correct?
So let's cross this one off.
And again that's the doubling issue that we talked about, correct?
And in the criminal trial -- you can look at that, refer to that chart -- you testified that actually the most favorable number for the defense for three contributors was 1 in 7, did you not?
So for the G11, the three contributor number at this time I reported as 1 in 7 to 1 in 1300.
I'm asking you about prior testimony under oath, Dr. Weir, the number you gave under oath in the criminal trial, in your cross-examination.
Double -- oh, we got that already. We got two double errors -- I'm sorry, we got one.
Now, for G1, G2 and G4, the PCR results, I'll represent to you that your criminal trial direct testimony was that that number, the most frequent pattern favorable to the defense is 1 in 240.
And that was wrong, correct?
(BY MR. BLASIER) Now, 31 -- PCR results in No. 31 -- 31 is again a stain from the Bronco, correct?
I'll represent to you that your criminal trial testimony was 1 in 4700. Your testimony here today is 1 in 6800, correct?
And for 31, you testified that for four contributors, in the criminal trial, the frequency most favorable to the defense would be 1 in 1300, correct? That's on your chart?
But it's -- the reason this number is wrong and the reason this number is wrong are both because of the doubling issue, right?
G12 and G14, the PCR results, I'll represent you testified at the criminal trial the most frequent occurrence was 1 in 6.
And your testimony here today is that that actual number is 1 in 11, correct?
So your estimate. All right.
And this one's wrong. All right.
And that's another double issue; is that correct?
Now, the G4, the RFLP results, your testimony here is that the number most favorable to the defense is 1 in 1 billion, correct?
For the two contributors.
The criminal trial you testified that that number was 1 in 1 million, correct?
Let's put a question mark there.
Now, the number for multiple contributors most favorable to the defense for G4 is 1 in 20,000, is it not what you testified to at the criminal trial?
So G1 and G2, the RFLP results, again we're talking about glove stains at the criminal trial.
You testified that the number most favorable to the defense was 1 in 6 million, correct?
That's right. It's the same -- it's the same difference because it's the same probes, that's really the same stain. They have almost an identical profile so that is a little misleading to separate them like that.
Let's cross it off. It's wrong, right? Correct?
And you have no idea where this number came from?
Okay.
Let's put a question mark over there.
And the most favorable number that you testified to in the criminal trial was for three contributors for this stain and it was 1 in 3,000, correct?
Double.
Now, number 9, the PCR results from the glove, I'll represent to you that you testified that the number most favorable to the defense in the criminal case was 1 in 14.
And you've testified here today that the number, the more accurate number is 1 in 29, correct?
The number is --- the number is calculated correctly, but it's not the number to present the 29.
(BY MR. BLASIER) Let's cross it off.
And the number that you presented in the criminal trial for three contributors for stain No. 9 as the one most favorable to the defense is 1 in 4; isn't that correct?
Okay. Cross it off.
And that's again the double issue for this one and the doubling issue for that one, correct?
So we had double over here.
Now, at your deposition you turned over some articles that you had written, correct?
And you were ordered to turn over a paper called or from the -- called the Promaga (sic) Conference, correct?
And in that paper you discuss the appropriate way to calculate some of these frequencies, correct?
(BY MR. BLASIER) That also discusses the procedure that you advocate for calculating percentages and mixtures, correct?
And you also presented a chapter from one of your books called "Genetic Data Analysis," correct?
Now, you actually wrote an article about the Simpson case, after the criminal case, that you didn't present yesterday; isn't that correct?
And that's an article which detailed DNA statistics in the Simpson matter from "Nature, Genetics, Volume 11, December 1995," correct?
Now, let me ask some questions about what those numbers don't take into account.
You would agree that these numbers assume that the evidence was properly collected, properly processed, there was no planting, no cross-contamination, that the actual results that the lab got were all correct, don't they?
Objection, argumentative, assumes facts not in evidence, beyond the scope, and in addition, the results have been admitted.
I'm not -- I think I need to answer carefully.
The calculations I do, start with the profile determination given to me by the forensic scientists. What -- Anything that goes on before then is beyond my knowledge or expertise.
(BY MR. BLASIER) Tell me if you agree with the following statement: It is essential that the integrity of DNA evidence with regard to collection, potential contamination or tampering be beyond doubt?
(BY MR. BLASIER) The integrity of your numbers depends on the integrity of the underlying evidence, does it not?
I don't see the connection.
I'm saying that the calculations follow from the profile. Anything else is not -- is not here. I've taken the profile, done some calculations. I don't know what else to say.
Sustained.
Excuse me, Mr. Blasier, you can develop, if you can, through competent examination and competent evidence any issues in this regard. That does not permit this witness -- I will permit examination of this witness on that. You may examine him as to his computations.
(BY MR. BLASIER) Now, Dr. Weir, you've testified here, I think you said it a couple of times, that the methods that you used here and the statistics that you have presented, are the most appropriate, correct?
Isn't it true, that it is actually your belief that the more appropriate statistics to present are not these, but likelihood ratios?
I -- we're talking words, Mr. Blasier. These are -- these are -- can be interpreted as likelihood ratios. The numbers are the same.
You said several times in the writings that you gave me, and in this article that I've talked to you about, that the more appropriate statistics, rather than these numbers, is a likelihood ratio, correct?
You are misquoting me.
The numbers are the same. They've been presented in court today as either frequencies or probabilities, which they are. They could equally have -- well have been described as likelihood ratios. The numbers, however, are identical.
Don't you point out in your paper that the requirements of likelihood ratios not being presented in the criminal trial emit that only certain numbers were given which ironically could have hurt the defense's case.
Let's go to the bottom of this chart.
And in that paper, you used as an example from the criminal trial, stain No. 31, which you've testified to about here that for four unknown contributors, the probability of the RFLP profile in Item 31 varies between 1 and 240 million and 1 in 2.7 billion, correct?
In your paper, you state that for four unknown contributors with respect to stain No. 31, the probability for the RFLP profile varies between 1 in 240 million and 1 in 2.7 billion, correct?
Well, I'm going to ask you about that.
Then you say further on your paper, that if you consider that -- rather than four unknown contributors, only have two unknown contributors, that the number gets reduced to 1 in 200,000, which would be much more favorable to the defense, correct?
That's right. That's consistent with what we've -- we were talking about earlier. The more contributors you have the less likely it is that those higher number of people have a particular profile.
The number of unknown contributors also makes a difference in terms of what a likelihood ratio number would be and a frequency number; isn't that correct?
Then you say in your paper that if the defense had been willing to concede that the blood in the Bronco was consistent or was from Mr. Simpson, then there would be a difference in the numbers that would reduce the frequency to 1 in 1,000, correct?
Before they put these numbers up, Your Honor, I would like to ask whether they are conceding in fact the blood in the Bronco belongs to Mr. Simpson. It appears to be the thrust of the question.
(BY MR. BLASIER) And the tone of that, would you agree, is that again if the defense had done it the right way, with your assumptions that you make in that article, they would have come up with a much more favorable number than the way they wanted to do it, correct.
I'm not sure what the question is. What the quotes you wrote out or are accurate. I think they speak for themselves.
The point you're making there is, is it not, that had the defense done it the way you thought was the appropriate way, with more reasonable assumptions, they would have come up with numbers much more favorable to the defense than they actually did?
Simply, as I said, if the defense had stipulated to one of the contributors, then the numbers change.
No. Well, excuse me. Yes, it was. It was invited and then it was reviewed before it was accepted.
And the peer review process involves someone else with qualifications similar to yourself, I presume, reviewing your work, checking your data, making sure that what you've stated in here is accurate; is that fair to say?
Now, tell me, Dr. Weir -- oh incidentally, by the way, none of these frequency calculations takes into account the possibility that an Asian person might have contributed to some of these stains, correct?
No, that's not what I said. The numbers presented are not based on Asian data. I can talk about Asian frequencies if you wish.
Now, this article that was peer reviewed, where you talk about Item 31, and a four probe RFLP match, where did you get your information that there was a four probe RFLP match for Item 31?
I don't remember at this time if it -- I imagine it came from the proceedings of the criminal trial if I -- if that's what I said.
I don't recall. I'm having trouble with this. I don't recall 31 having an RFLP or am I wrong?
You're correct here that 31 did not have any RFLP results, but your article that was peer reviewed said it did, correct?
All the calculations I did involving DQ Alpha in my hotel room were wrong.
I was being -- I guess I was being lazy on Tuesday.
You have to tell me what you've done there... This -- excuse me. You have to tell me what you've done there.
I don't think I want to play that kind of a game. But there are several people who do these calculations just as well as I do.
That's a thousand-fold difference, isn't it?