Now, the section of the National Research Council report that we are talking about on page 59 has the following language. It says: "If a suspect's pattern is found within the mixed pattern, the appropriate frequency to assign to such a match is the sum of the frequencies of all genotypes that are contained within."
And the term "Genotype" doesn't refer to the genetic characteristics of three individuals either, does it?
No. So the term "Genotype" refers to the genetic characteristics of a single individual; isn't that right?
That's right, but of course the language of this report, if we stuck to that, that would make no sense, so I interpret "Genotypes" here to mean genotypes to the contributors of the mixed stain.
But that is your interpretation; that is not what the National Research Council says, is it?
The National Research Council is reporting mixtures. We couldn't do a mixture analysis ignoring the fact that it is a mixture, so I take this as just loose language.
Dr. Weir, I would like you to take a look at this chart that has previously been marked chart 410 that has been prepared by the Defense, and I think you will see on this chart there is a list of numbers that are summed.
All right. Now, let's take a look at chart no. 409, please. Now, Dr. Weir, are you looking at chart 409 which was previously prepared by the Prosecution?
All right. Now, I would like you to look at this--the number that is listed under the column headed "Frequency," this number .15 percent.
And what it actually is, is the product of the frequencies of two genos, isn't it?
And the reason that that number is so small, that is the .15 percent, is that you have taken two--two numbers and multiplied them together, taken two percentages and multiplied them together to get a smaller percentage? Isn't what you have done?
All right. Now, does the National Research Council report say anything about taking the product of two frequencies before you sum up these--
No, no, but that is implied by the language. The frequency of the genotypes couldn't mean anything else.
So your conclusion is that it couldn't mean anything else? That is the way you read it?
Certainly we are talking about a mixture. To talk about an analysis of a mixture ignoring the mixtures would make no sense and I don't believe that that was what was intended.
--the NRC report meant or is this a new interpretation you arrived at only recently?
I had never thought about this until quite recently. I have always done it the correct way.
Well, what perplexes me, Dr. Weir, is that you have prepared several different versions of a report in this case.
And in the report with Dr. Buckleton you use likelihood ratios, a different kind of statistics--
All right. Let me rephrase the question. You use likelihood ratios to provide a measure to the extent of which the evidence favors one scenario over another; is that correct?
The complete analysis must be in that framework. Of course a part of that calculation is the part where we are discussing at this point. This is the first part and in the complete and appropriate analysis.
All right. And then on May 31st you produced another report labeled a draft. This one single page authored by yourself; is that right?
All right. And this report also used likelihood ratios to characterize the value of mixtures?
Oh, as does my current report. The phrasing is still there, but my current report is focusing on this component of the likelihood ratio.
Uh-huh. And in fact the report says likelihood ratios are essential in interpreting the evidentiary value of mixed stains, doesn't it?
Yes. That is true, yes, whether that is single or mixed and of course we have had them throughout this case. Every number that is on the board is in fact giving us a likelihood ratio. The numbers I have discussed this morning for the mixture stains are giving us likelihood ratios by inverting them.
Okay. Are you planning to tell the jury that those frequency numbers are in fact likelihood ratios?
Okay. Then there was another report that was produced just this week with the date of June 20th. Are you familiar with that? That was June 20th is actually Tuesday.
There is a June 20th report. On Tuesday we got a report. This report also used likelihood ratios; is that right?
Well, all the reports have likelihood ratio language in them. All the reports also calculate frequencies of multiple contributors to a mixed stain. I'm not sure I understand what the distinction is you are trying to make.
Okay. And in the report of June 20th you are quite critical of the National Research Council's method for dealing with mixtures, are you not?
Right. In fact, I see the sentence that reads: "In particular the method of assigning frequencies for mixtures suggested by the NRC report ignores the type of the profile of the person or people of interest and therefore has no probative value." Do you remember that sentence being in your report?
Yes, but that is a separate issue of course that is talking about the enumerator of the likelihood ratio and that criticism of course is still with us. The--I think we are trying to confuse two issues here. Whether or not we give likelihood ratios is in a sense immaterial, but if we were to, we are comparing frequencies under two scenarios. For example, contributors being known or not known. There are two things. I believe if you merely give the bottom line, you are leaving out the alternative, so regardless of how you calculate the bottom line.
Isn't it true, Dr. Weir, that when you were criticizing the National Research Council report in your--in your June 20th report--
--you were referring to the interpretation of the National Research Council method for mixtures that has been advocated by the Defense in this hearing?
I don't think so. My criticisms are whether or not they--we are comparing frequencies under alternative scenarios.
Yes, I'm just--I'm just flicking through pages here. Maybe could you help me and refer me to the page.
Now, am I right that the third sentence of that paragraph reads: "Simply adding the frequencies of all possible contributors to the mixture (NRC 1992) Ignores the essential nature of a mixture"?
Now, when you wrote this sentence you were assuming, were you not, that the NRC approach involved simply adding the frequency of all possible contributors to a mixture, weren't you?
I think that sentence would--would criticize it if that was the intention. I think, though--
I think, on reflection, I was being unfair to the NRC. I don't believe they would have meant a single contributor. That would have made no sense.
All right. And so the--the interpretation of the NRC report which is reflected in the report of June 21st is not the interpretation that you now would adopt? Is that what you are saying?
We are getting hung up with words here. The correct interpretation of any evidentiary stain is to compare the frequencies and their alternative scenarios. It's simple in a single stain because the likely--the chance that the frequency of the stain under one scenario is just one, it is a hundred percent, so in essence the frequency is hidden from us. And then we calculate the frequency for the unknown and we don't run into any need to examine what we are doing. We can--we can and we have seen the possibility of sloppy thinking, because we have ignored the alternative. For the mixtures we need to consider both alternatives.
However, one of those alternatives involves multiple unknown contributors and there is only one possible way for calculating the frequency of multiple contributors, regardless of what the NRC said or meant to say or is purported to have said. It is not a question of interpretation. Finally, it is a question of how do you calculate the frequency of multiple contributors, and there is only one way to do that.
And you are not relying on the NRC report for the basis for your methods in any case, are you?
The report was published in 1990. You have hit on the points. Here is a report which ignores the literature. It is amazing to me that in this and many other instances they have chosen to ignore the correct procedures. Why they didn't cite this paper, I don't know.
Right. So the very paper on which you were relying was simply ignored by the National Research Council?
All right. All right. Now, this paper by Dr. Evett, the method that it describes is for characterizing mixtures in situations where we know how many contributors there are, is it not?
Yes. I think so. That is--well, I believe all the--all our calculations assume a number of contributors, yes.
Right. Now, but in this case we cannot tell by looking at most of the mixtures how many contributors there are, can we?
Oh, I think the evidence is overwhelming that there were two contributors when we have seven RFLP loci and seven PCR loci and I have at most four alleles. I can't imagine any other possible interpretation.
So does this evidence tell you--can you tell just by looking at the genetic evidence the probability that it is two persons who contributed versus three?
I'm not sure about that. I would work--yes, I could work out the probability of that event conditional on each of those two assumptions.
Right. But in order to--in order for you to compute the probability of getting these mixtures, assuming that there are either two people or three people, you would have to make some assumptions about the a priori probability of there being two or three contributors, wouldn't you?
No. Well, excuse me. To give a posterior probability of the time analysis, I think the analogy is flawed; however, it ignores what we have been doing throughout the whole trial for a single stain. We compute the frequency of single contributors because all the evidence suggests there is a single contributor. No one has ever suggested we should compute the evidence of a single stain as though there were multiple contributors. All the evidence suggests there are not multiple contributors. I find the argument somewhat vacuous.
Okay. The question is, in order for you to compute the probability of getting these mixtures, assuming there are either two or three people, would you have to make some assumption about the prior probability of there being two or three people?
So my answer is certainly no to that. Given--you read that again, you will see that is a conditional probability you have just asked me about.
All right. Suppose I framed it as a posterior probability. In order to get the posterior probability being two people versus three people--by posterior probability we mean the ultimate issue of there being two or three--we have to make some assumptions about the likelihood there are two or three there initially before we look at the evidence?
We are not making posterior probabilities about anything in this trial, single stains or mixture stains or any other issue before us. We have no priors to put in to end up with posteriors.
When you say the evidence is overwhelmingly in favor of there being two contributors to a stain--
Yes, there is. There are seven RFLP loci, each of which has at most four alleles. There are seven PCR loci--well, depending on the stain--each of which has at most four alleles. How could it possibly be three contributors? It is beyond imagination.
But in order to draw conclusions about the relative likelihood, you have to make some assumptions drawing on other evidence, don't you?
The assumption I choose to do these frequencies is that there are either two or three contributors based on the evidence. There are two contributors. I have chosen to do it for three. I could have done it for four. I see no point.
What about stain no. 29 where we are talking about the stain on the steering wheel?
Can you tell by looking at that stain whether there are two or three contributors?
I can see no evidence there which suggests three, but you will need to give me some time to look at the RFLPs. They are there. So we have only the PCRs and there are three alleles at all those loci. That is six loci. Excuse me. Six loci have three alleles. There is nothing there which would suggest three contributors.
All right. And you say--but you can't determine, based on those three alleles, whether there are two people or three people there, can you?
There are three alleles. We know that each person has two. People may share an allele, so two people may have two alleles, one of which is shared, which would end up with three. I can't imagine why you would invoke three people when there are only three alleles. There may be three people. There may be six people. There is nothing there which would require me to assume more than two.
All right. And so based on the genetic evidence alone there is no basis for determining whether there is two or three or four?
The evidence is that there are only at most four with here only three alleles. The polymarker had four--I think--I would have to look. I think only three alleles. So the evidence says there are three alleles. That means more than one person. It does not say anything about there being three people.
Uh-huh. And so you can't tell whether there are two versus three based on this evidence alone?
All right. Mr. Thompson, you have to understand I'm not making this determination in a vacuum. I assume that we have two victims at the crime scene. I'm taking into consideration that there appears to be only one set of footprints. I mean, this is not a determination made in a vacuum. Proceed. Proceed.
Now, Dr. Weir, the Defense in this case has taken the position that numbers need to be presented to characterize the value of evidence showing consistency between the DNA profile of an individual and a mixed stain. Do you agree that numbers are necessary to show the value of such evidence?
My opinion has changed. I think if you had asked me two years ago, I would have agreed. We are now at the point that the numbers are beyond belief. We have got so much genetic evidence, there are so many loci types, some of these items have 11 RFLP loci and 7 PCR loci. There is no number that makes any sense, that is beyond the belief--beyond the experience of anyone to try and interpret. I believe we have passed beyond the point where we should run around trying to make up numbers acceptable to everybody. The evidence of a match at this many loci is compelling, so I think once when we got to the point we are presently at, 7 RFLP loci, 7 PCR loci, the numbers are not necessary.
KEY QUOTEWhat about for evidence like that on item 29? Would the comments you just made apply to that stain as well?
Well, my comments apply to seven of each type. On item 29 there are many fewer loci scored.
The events of a match is not so astonishing and it is indeed helpful to have a number.
And so I believe the intention of the Prosecutors initially was to present no number in connection with the mixed stains but merely to present the profile frequencies of individuals who were being compared to the stain. Would you agree with me that that would be a misleading procedure for characterizing the value of stains like no. 29, 305 and the other Bronco stains?
Well, that is--I would need to examine each item separately. It would depend on the number of the loci.
Let's start with no. 29 then. Do you think it would be misleading just to present the individual profile of individuals being compared to the stains, such as Nicole Brown Simpson, and not compute a special number that refers to the likelihood of her matching with the mixtures?
It is not misleading, no. The evidence remains that there was a match between the evidence and known sample. That is not misleading to say that.
It is helpful to interpret that evidence to attach numbers under alternative scenarios.
Uh-huh, but the question is would it be misleading to present just Nicole Brown's individual profile frequency and not do some computation regarding the likelihood of her matching a mixture?
That is not the question you asked me. I don't know why we are computing Nicole Brown's profile frequency. Why are we doing that?
Well, let me ask you the question this way: Would it be misleading to do what the Prosecutors were initially advocating in this case, which is to present no number on the mixture, per se, but just to present the profile frequencies of individuals who were being compared to the mixture? Would that be misleading?
I don't know how to answer that. I'm not--I don't understand the question. Umm--there was a mixture. You either attach a number to it or you don't, but there was the evidence of a mixture. You either attach a number or not. It is not misleading not to attach a number. It is helpful to attach a number. Calculating one of the--one of the matching known frequencies is beside the point. It has nothing to do with the mixture, so I don't understand.
Well, let's suppose that the frequency of Nicole Brown's DNA profile on DQ-Alpha and the polymarker systems ranged from 1 in 2500 to 1 in 26,000. Do you understand that assumption, a relatively rare profile across--across DQ-Alpha and the polymarker loci?
Right. Do you think it would be misleading for the Prosecutors to give a number like 1 in 26,000 to Nicole Brown's profile and then to say through testimony that Nicole Brown Simpson's profile is consistent with stain 29 and give no further number?
I'm not sure I would say misleading. It doesn't--it reflects a fallacy of trying to give any of this evidence outside the appropriate framework. Unless you give frequencies attached to items of evidence, and their alternative scenarios, you cannot make a determination which in fact was the most likely scenario. So anytime you try and concoct these methods of presenting evidence outside the appropriate way of doing it, you will run into problems. I mean, you or anybody.
Uh-huh. And by your calculations, the probability that two randomly chosen individuals, one Caucasian and one African American, would--would match with stain no. 29 is 1 in 71; isn't that right.
Right? Which is considerably higher than Nicole Brown Simpson's profile frequency, isn't it?
All right. All right. And so it would be incorrect, wouldn't it, to use Nicole Brown's profile frequency as an index of the value of the fact that she was consistent with that stain, wouldn't it?
All right. Okay. Now, so it sounds like the real--the real issue here is over methods for doing these mixture calculations; isn't that--and we basically agree a number is needed to characterize a mixture, and our disagreement is how to compute that?
Okay. All right. Would you agree, Dr. Weir, that in order to do the kind of calculations that you are doing on the exhibit currently on the elmo, which I believe is no. 410--
409. In order to do the calculations on chart 409, you need to make certain assumptions?
And one assumption you need to make is with regard to the number of contributors to the stain?
Now, does that assumption, with regard to the number of contributors, does that assumption have to be made to do the calculations in the manner reflected on chart 410, which is the Defense's approach to the calculation?
It doesn't, but that is not a relevant fact. We have no choice. It is not a question of taste or opinion. If we want to estimate the frequency of the contributors to the mixture, we must take account of the number of contributors. I'm not sure how I can say that any more simply. It is not rocket science here.
Now, another assumption that your approach makes is that all alleles of all contributors can be detected; is that correct?
That's true. I have had to assume that I am not an expert in the molecular biology. I take what the forensic scientists approach.
Does that assumption need to be made to make the calculations reported by the NRC counsel I'm chart?
Absolutely. You can't do anything until you have a starting point which is at least found. Now, I will continue, because there are cases where we know there were two contributors, but we see less than four alleles at a locus. It shows up more clearly in the RFLP situation where some of the alleles appear either or hidden or to be not seen.
Now, based on the calculations done on this chart, Dr. Weir, would it be correct to say among Caucasians the percentage of people who have genotypes consistent with alleles observed in stain 29 is approximately 45 percent?
That is a completely misleading statement and I don't even believe it is true. "Consistent" means you can account for the stain with the people listed. You cannot account for the stain by adding up those single contributors. The first item does not account for alleles 1.2 and 4. If you--that statement you made, and I listened very carefully, I think comes very close to being dishonest.
Yes, it is incorrect scientifically because it is not--I take "Consistent" to mean explaining the data.
All right. Let's--let's instead ask the question this way: Would it be correct to say that among Caucasians the percentage of people who have genotypes that would be included in the stain observed in stain 29--included in the alleles observed in stain 29 is approximately--let me--let me start again. Based on this chart, no. 410, would it be correct to say that among Caucasians the percentage of people who have genotypes that would be included in the set of alleles observed in stain 29 is approximately 45 percent?
That statement is correct, but it is misleading in the sense it gives us no basis for interpreting the evidentiary value of this mixed stain. It is a correct statement, but I don't think it has any relevance.
KEY QUOTENow, when you say "Misleading," what you mean, I take it, is that you think it might give the jurors the wrong idea about how to think about this evidence?
I mean it is misleading in that it gives us no basis for interpreting the evidence of matching between known contributors and a mixed stain. It gives us no foundation to making--drawing any conclusions numerically.
Well, suppose that I, upon learning that Nicole Brown Simpson's genotypes are consistent with those found in stain no. 29, was curious to know what percentage of people in the population have genotypes consistent with those in stain 29. Isn't 45 percent the number that I need to know to respond to that--
You might want to know that, but that is not a question that we would need to be addressing here. It is not relevant. I mean that is an interesting question, I suppose, but why would you even ask it? The evidence is that there are two known contributors. They match the mixed stain. Is that--how often does that happen coincidentally?
If you want to talk about Nicole Simpson, you need to consider the scenario of Nicole Simpson versus another unknown versus two unknowns.
Okay. So if I understand your argument correctly, you are arguing that the way the jury should think about the problem or the way anyone else should think about the problem is in terms of the likelihood of getting a pair of individuals that would match rather than just a single individual who would match? Is that your position?
Excuse me. I recent that completely. That is not just an argument. That is the only conceivable way of attaching a frequency to the match of two known contributors in a mixed stain. How could you possibly do anything else? And to characterize it as just an argument I think is unfortunate.
Doctor, you keep using this phrase "Known contributors," as in two known contributors.
All right. But we don't--we don't know who the contributors are of any stains in this case based on the genetic evidence alone, do we?
Well, that is what we have to--what the jury eventually will have to decide, are these two known people the contributors or are two other people so we are comparing those two known people with two unknown people. So that is not an assumption; that is what the trial is about, to make decisions on the origins of these stains.
All right. To answer questions about the number of people in the population or in some population that would have genotypes that would be included in a mixture, such as no. 29, is it necessary to make assumptions about whether all alleles of all contributors to the mixture have been observed?
I have answered and on the RFLPs we don't--we--my calculations show that we allow for unseen bands.
Well, what about--let me restate the question and I believe it is a different question this time, your Honor.
If--if the question that I want to have answered is--is what percentage of the population would have genotypes consistent with the observed alleles, isn't it true that I don't have to make any assumptions at all about whether all alleles of all contributors have been observed?
I think I understand the question. The answer is yes and that is what I have done. My calculations assume these are the only alleles in the--in the mixed stain.
But to do it--to do it the way reflected on the Defense chart where you are simply summing, and--and asking the question what percentage of the population has genotypes consistent with the observed alleles, we need make no assumption about whether all alleles have been observed; isn't that correct?
You don't need to make any assumptions but then you don't end up with a meaningful answer.
Right. Whereas under your approach assumptions do need to be made about whether or not all alleles of all contributors have been observed?
That is correct for the mixture and the single stains, is that the foundation of the whole forensic uses of DNA we have alleles seen, we calculate frequencies for them. If one was to object that that assumption, one might as well hang up their hat and go home because none of this would be possible. Of course we assume the number of contributors is consistent with the evidence.
Okay. So there are two assumptions that are being made under your approach which need not be made--
There are two assumptions being made there that need not be made to use the kind of numbers that are presented in the Defense chart, one being the number of contributors and the other being the connection between the observed alleles and the contributor's allele?
Yes. Okay. Now, doctor your expertise is in drawing statistical inferences from genetic data; is that correct?
One of the issues that statisticians like yourself are most concerned with is what is and isn't a logical from a scientific finding?
And I often hear statisticians use the term "Permissible" as in this is a permissible--
Permissible, as in this is a permissible inference or this is not a permissible inference. Is that a common expression in your field?
Oh, okay. What expression would you use to talk about an inference which is illogical and inappropriate?
Wrong. Okay. Now, I gather when statisticians say an inference is wrong, it means it is illogical or fails to follow from the premises?
Okay. And the question about which inferences are correct and incorrect, wrong or right, when interpreting scientific data, is one of the major questions that statisticians concern themselves with?
I don't see statistics as being consumed with giving right or wrong answers. Statistics is concerned with teasing out the information and sets of data, how results are often traced in terms of estimates or results of hypothesis testing. Statistics is not a black and white, right or wrong, yes or no kind of a science. It is a science of interpreting data, so I'm having trouble with understanding your characterization of statistics.
But part of what statisticians do is determine what is and what is not a reasonable statistical inference from data; isn't that correct?
Counsel, the issue is which of these two methodologies am I going to adopt for use in front of this jury.
Which makes sense both in terms of the science that produces these results and the math that is applied to it.
Dr. Weir, when drawing conclusions from a scientific tests, should an expert witness in a criminal case rely only on the scientific data or is it permissible for the expert to take into account other evidence in the case?
These--this is such a general question I don't understand it. I will say yes, but I'm not fully understanding what you are meaning.
Well, suppose, for example, that an expert sees a scientific test result that has two possible interpretations and one interpretation is consistent with guilt and the other interpretation is consistent with innocence. Would it be permissible for the expert to consider the overall strength of the Prosecution's case when deciding which scientific interpretation was correct?
Would it be permissible for a statistician, in determining which statistical procedure is appropriate or inappropriate, to take into account the strength of the Prosecution's case?
Dr. Weir, have you heard the term "Bootstrapping" when that term is used to refer to allowing one set of evidence to influence a supposedly independent judgment?
Now. Yeah. Have you heard the term "Bootstrapping" used in the scientific arena, as opposed to the legal arena?
Right. Now, I know your report refers to a statistical procedure known as bootstrapping. Have you heard an additional meaning of that term "Bootstrapping" in scientific circles?
It is really not. It is a complete waste of my time so far, this whole discussion about bootstrapping.
KEY QUOTEWell, using the method that you are advocating, you need to make assumptions, and would you agree with me that the numbers that are produced by your method vary depending on the assumptions that you make?
For example, the numbers that you produce would vary depending on whether you assume there were two contributors or three contributors?
I also have tables for single stains where I have assumed a single contributor. I have answered that several times.
All right. And the numbers that you compute would also vary, depending on whether you assumed that all alleles of all contributors are being observed in the stain?
Okay. And so under some assumptions you get numbers that are much bigger and more impressive than under other assumptions, don't you?
And have you noticed, Dr. Weir, that when you compute numbers under assumptions that are consistent with the Prosecution's theory of the case that you will get numbers that are larger and more supportive of the Prosecution's case?
Is it the case, Dr. Weir, that if you compute numbers under assumptions consistent with the Prosecution's theory of the case, that the numbers that are produced by the method you are advocating are more supportive of the Prosecution's theory of the case?
Mr. Thompson, if it is of any benefit to you, I understand the two assumptions that underlie Dr. Weir's testimony and the calculations that he makes and the fact that you will get different results depending on what the assumptions are. I understand that. I understood that an hour ago.
Let me ask you a few questions, Dr. Weir, about what your report says concerning item no. 78, and I think if you look--if you look in your report at page 25--
It contained a number of alleles which were consistent with Nicole Brown Simpson on RFLP tests, along with a few that were consistent with Ronald Goldman.
Yeah. I was asking you if it is correct to say that item 78, the stain on Mr. Goldman's boot, contains a number of alleles consistent with Nicole Brown Simpson together with a few that are consistent with Ronald Goldman?
Okay. Mr. Goldman also has a few alleles that are not observed in item 78; is that correct?
Yes. This is an MS43 allele not in the mixture and two YNH24 alleles in Goldman's profile not in the mixture.
Now, I see in your report on page 7 that you say Ronald Goldman is not excluded from that stain?
Okay. All right. But Ronald Goldman does have some alleles that are not in the stain, doesn't he?
Well, now you are getting into an item which is beyond my expertise and I believe we've had testimony from Cellmark as to the reason they call the matches and non-matches and exclusions. My analysis must start with the forensic scientist's determination and they said there was a failure to exclude based on the molecular biology and the description of the staining patterns which is beyond my expertise to interpret. I will accept their interpretation. I can't do anything unless I have a starting point. The starting point is the determination made by the forensic scientist. Given that starting point, my analysis follows. So whether or not alleles match or not, are found are not, that is where I start.
Okay. Well, let's start--let's start with your chart on page 25. Would you degree with me that your chart shows Ronald Goldman having some alleles that are not observed in item 78?
So would you agree that Cellmark must have assumed that some alleles of some contributors at some loci to item 78 were not observed?
I think I have already said that. My starting point is the determination of non-exclusion. All these other things will follow from that.
Okay. So this is one of those cases where assumptions about whether all alleles have been observed come into play? Would you agree?
I have said that several times repeatedly and I'm going to say it again. Maybe we could start the question. I'm going to take the determination by Cellmark that there were two contributors, neither brown nor Goldman are excluded, and my calculations will proceed from that point. All these other things are details of that statement.
Now, did you make--when you were doing computations for item 78, did you make the same assumptions with regard to alleles showing up in making your statistical computations that Cellmark made when determining the match?
I don't know how else I can say it, I'm sorry. That was the same question as before. Cellmark has determined a match, they have not excluded two people. I start from that and proceed. I don't know how else I'm going to say it.
All right. And would that--does that mean then, that on those loci where Mr. Goldman's alleles do not show up in the mixtures, you assumed there were missing alleles?
And on those loci where Mr. Goldman's alleles or those matching--alleles matching Mr. Goldman were found, you assumed there were no missing alleles, didn't you?
My assumptions are that Cellmark has declared a match with these two stains, not excluded people. The bands showing in Goldman I'm going to use in my analysis, whether or not they show in the mixtures, because he has not been excluded from the mixture.
The analysis is that there was a mixture consistent with these contributors, Goldman's bands are not showing in the mixture YNH24. That is not my assumption. That is what Cellmark told me their determination was. My analysis reflects that determination. I assume nothing.
So, for example, on locus G3 you are proceeding, or your analysis proceeds from the premise that there are unseen alleles in the mixture, that there are--not all--not all alleles of all contributors are seen in the mixture, aren't you?
There is no need to. Under the scenario of known contributors we have all bands accounted for. Under the scenario of unknown contributors, we need to work with the four unknowns. We need to look at this entire package. The package is consistent with there being two contributors. I have therefore conducted an analysis on the basis of there being two unknown contributors, which would result in those four alleles at MS1.
So you made that--you made the assumption of unseen alleles where that was necessary and you didn't make it where it was unnecessary; is that right?
I haven't assumed unseen alleles. I say there are alleles I have seen. I'm going to assume there were two contributors. I'm going to compute from two contributors.
Whether your calculations proceeded on the basis that there were missing alleles or not at a particular locus depended entirely on whether that assumption was necessary to make Mr. Goldman match; isn't that true?
My calculations have nothing to do with matching. That is the top line of the likelihood ratio. I'm doing two unknown contributors. Two unknown contributors. How do two unknown contributors give those two alleles?
What is your independent scientific basis for assuming that there were missing alleles at YNH24, for example, but not at MS1?
I have no outside evidence other than what Cellmark told me. All of my analyses are based on what the forensic scientists have told me. If they told me this, then I will proceed accordingly. It is beyond my expertise to do otherwise.
And had they assumed--had they made the opposite assumption, for example, assuming there were missing alleles at MS1, but not at YNH24, Mr. Goldman would have been excluded, wouldn't he?
Dr. Weir, isn't your logic in drawing conclusions from item 78 circular, in that you are basing your conclusions on premises that are consistent with the Prosecution's theory of the case?
Now, Dr. Weir, in your report you mention hypotheses that can be tested by the data and you label them c and c prime. Is it called c prime when you put--
Contact between the known person and the location of the evidentiary stain. We can say hypothesis or scenario. I'm not sure what the best word is. I think I would use them interchangeably.
Okay. And so your report suggests that two hypothesis that are to be distinguished by the genetic data are contact and no contact?
Okay. And so, for example, with regard to stain 29, if we were interested in whether Nicole--Nicole Brown Simpson's blood was in that stain on the Bronco steering wheel, a no contact would mean known of her blood had gotten into the Bronco? Is that what--into that stain. Is that what that would mean?
I'm not sure--I'm not sure. We need to be careful. The contact there would be between Nicole Brown Simpson and the steering wheel. Somehow under that scenario her blood went from her body to the steering wheel. That is contact. It not might not mean with her physical self, but some means of contacting her blood--contact of her blood with the steering wheel, whatever the agent was for making that contact.
And--and are you--are you planning to testify that your frequencies represent the probability of this evidence under the assumption of no contact?
All the numbers on this chart are frequencies and frequency and probability I use interchangeably. All these numbers we've held in the entire case are the probabilities of the evidence under the situation of no contact. In other words, a random person. Someone other than the known, than the stated person, having contributed that.
Are you assuming, Dr. Weir, that a coincidental match is the only way in which the evidence we have observed can arise under the--under the hypothesis of no contact?
So you are assuming that there other ways that this evidence could arise that are not reflected in your numbers?
I'm not assuming anything. I have made some calculations accounting for the evidentiary stains from unknown people. I'm saying nothing else.
A frequency is the frequency of the evidence given known or unknowns. If you start taking ratios of frequencies--I'm sorry if you don't like the term, but that ratio of frequencies is a likelihood ratio.
And when you testify that there is--when you put a--when you testify that there is a 1 in 71 chance of observing the pair of contributors--a pair of contributors--well, there is a 1 in 71 chance that a pair of contributors at random could have left the stain on item 29--
I don't believe it is easy for you to fall into that trap at all and I'm certainly not going to fall into it by you leading me into it.
I was actually leading you into another trap, sir, but let me proceed. Well, you expressed the term--you used the frequency 1 in 71 in connection with the stain on the steering wheel of the Bronco. Could you explain just exactly what exactly that 1 in 71 means?
That is the frequency or probability and I will use them interchangeably. That is the probability of seeing that mixed profile from two unknown contributors of specified ethnic origin.
Okay. When you compute the probability of seeing that profile under the assumption of no contact, you are assuming that the only way we would see that profile under the assumption of no contact was a coincidental match, aren't you?
I don't--don't think that is correct. I've calculated the frequency with which--of getting this profile from two unknown contributors. That actually doesn't say anything about contact or not. It is the contact refers of course to the known people, the people not excluded. You better restate your question I think.
Well, well, I guess the trouble I'm having is with the two hypothesis of contact, no contact, and your assertion is that the--the statistics that you are presenting give us the conditional probability of this evidence under no contact, isn't that--
Okay. But we could get the observed evidence in a number of ways other than simply the ways taken into account in your frequency calculations, couldn't we?
I--I--I'm very, very careful at the first page of my report to say my entire analysis is concerned with the genetic statistics. It assumes matches when declared to be true. It assumes profiles when declared to be accurate. It is--it would be wrong of me to make statements outside that context.
But if you are not taking into account anything other than coincidental match frequencies, isn't it also wrong to tell the jury that those frequencies characterize the probability of this evidence under the assumption of no contact?
I don't believe so, no, not in the way--not--if--if we are careful, I have been careful in my writing, if we are carefully orally, I don't think there will be any confusion.
You mentioned in direct examination that you work as a consultant for Cellmark diagnostics?
You also said you are a consultant with a number of other forensic laboratories doing analysis of their databases?
No, I didn't say that. If I did, I misspoke. We have similar contractual arrangements with other companies.
Okay. Now, in all cases where you act as a consultant for these forensic labs, does the money flow to the university or does it flow to you?
Several state agencies. Not my own state. South Carolina, Minnesota. I'm hoping to get some money from Illinois, although it hasn't come yet. Broward County in Florida.
Genetic design we have a contractual arrangement. I think I've received a little money from them some years ago, but I'm not sure. Some of that was sort of--they were acting as a conduit from a third party. There may have been a little money from them a couple of years ago.
Okay. Now, I see in your grant that you received a grant from the Department of Justice?
No. I have a twelve-month salary. I receive nothing directly or indirectly from that grant.
It is contained, as you can see from the application, to compensate a person working for me on these issues.
When you applied for this grant from the Department of Justice, were you given any promises of access to data from the FBI that other scientists would not have?
That is a two-part question. I was promised access to the FBI data. I know nothing about other people.
No, I'm not. In fact, I'm losing money. The county is not paying enough for my meals each day.
With regard from the beginning to the end, there will be not one dollar comes to me.
Flowing the other direction. I think that is very public-spirited of you, sir. One moment.
That statement is correct, but it is misleading in the sense it gives us no basis for interpreting the evidentiary value of this mixed stain. It is a correct statement, but I don't think it has any relevance.
That statement you made, and I listened very carefully, I think comes very close to being dishonest.
I resent that completely. That is not just an argument. That is the only conceivable way of attaching a frequency to the match of two known contributors in a mixed stain.
It is a complete waste of my time so far, this whole discussion about bootstrapping.
My opinion has changed. I think if you had asked me two years ago, I would have agreed. We are now at the point that the numbers are beyond belief... I believe we have passed beyond the point where we should run around trying to make up numbers acceptable to everybody.