Oracle Corporation et al v. SAP AG et al

Filing 836

Declaration of Scott W. Cowan in Support of 835 Memorandum in Opposition, filed bySAP AG, SAP America Inc, Tomorrownow Inc. (Attachments: # 1 Exhibit A, # 2 Exhibit B, # 3 Exhibit C, # 4 Exhibit D, # 5 Exhibit E, # 6 Exhibit F, # 7 Exhibit G, # 8 Exhibit H, # 9 Exhibit I, # 10 Exhibit J, # 11 Exhibit K, # 12 Exhibit L, # 13 Exhibit M, # 14 Exhibit N, # 15 Exhibit O, # 16 Exhibit P, # 17 Exhibit Q)(Related document(s) 835 ) (Froyd, Jane) (Filed on 9/9/2010)

Oracle Corporation et al v. SAP AG et al

Doc. 836 Att. 1

EXHIBIT A

Dockets.Justia.com

BRUCE SPENCER, PH.D. June 4, 2010 HIGHLY CONFIDENTIAL - ATTORNEYS' EYES ONLY
Page 1
UNITED STATES DISTRICT COURT NORTHERN DISTRICT OF CALIFORNIA OAKLAND DIVISION ORACLE CORPORATION, a Delaware corporation, ORACLE USA, INC., a Colorado corporation, and ORACLE INTERNATIONAL CORPORATION, a California corporation, ) ) ) ) ) ) ) ) Plaintiffs, ) ) vs. ) No. 07-CV-1658 (PJH) ) SAP AG, a German ) corporation, SAP AMERICA, ) INC., a Delaware ) corporation, TOMORROWNOW, ) INC., a Texas corporation, ) and DOES 1-50, inclusive, ) ) Defendants. ) ________________________________)

VIDEOTAPED DEPOSITION OF BRUCE SPENCER, PH.D. _________________________________ FRIDAY, JUNE 4, 2010

HIGHLY

CONFIDENTIAL - ATTORNEYS' EYES ONLY

REPORTED BY:

HOLLY THUMAN, CSR No. 6834, RMR, CRR (1-427339)

Merrill Legal Solutions [phone]

3c2f762b-a991-489c-93b9-1d08b41d0163

BRUCE SPENCER, PH.D. June 4, 2010 HIGHLY CONFIDENTIAL - ATTORNEYS' EYES ONLY
Page 92

TEXT REMOVED - NOT RELEVANT TO MOTION

11:25:21 11:25:24 11:25:35 11:25:39 11:25:43 11:25:46

15 16 17 18 19 20

Q.

So then how can you say this is a large

margin of error? A. In the statistics that I do, having a 25

percent error in what you're trying to measure tends to be a large error. Usually we try to get

more precision than that, in my experience.

TEXT REMOVED - NOT RELEVANT TO MOTION

Merrill Legal Solutions [phone]

TE 762b-a991-489c-93b9-1d08b41d0163

BRUCE SPENCER, PH.D. June 4, 2010 HIGHLY CONFIDENTIAL - ATTORNEYS' EYES ONLY
Page 95

TEXT REMOVED - NOT RELEVANT TO MOTION

11:29:21 11:29:22 11:29:27 11:29:29 11:29:30 11:29:31 11:29:32 11:29:34 11:29:35 11:29:36 11:29:38 11:29:41 11:29:42 11:29:43 11:29:44 11:29:45 11:29:46 11:29:48 11:29:51

6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24

MR. PICKETT:

Q.

Do you understand that

there is a burden of proof in the trial? MR. WILKES: THE WITNESS: is, yes. MR. PICKETT: burden of proof is? MR. WILKES: THE WITNESS: MR. PICKETT: Objection. Form. Q. Do you know what that Objection. Form.

I would expect that there

Not exactly. Q. Do you know what

confidence level the jury must have in order to make a determination in the case? MR. WILKES: THE WITNESS: MR. PICKETT: information on that? MR. WILKES: THE WITNESS: Objection. Form. Objection. Form.

Not exactly. Q. Do you have any

Well, I would ask an expert

before I wanted to trust my own judgment on that.
TEXT REMOVED - NOT RELEVANT TO MOTION

Merrill Legal Solutions [phone]

3c2f762b-a991-489c-93b9-1d08b41d0163

BRUCE SPENCER, PH.D. June 4, 2010 HIGHLY CONFIDENTIAL - ATTORNEYS' EYES ONLY
Page 97

TEXT REMOVED - NOT RELEVANT TO MOTION

11:31:14 11:31:16 11:31:18 11:31:20 11:31:20 11:31:22 11:31:24 11:31:25 11:31:27 11:31:28 11:31:30 11:31:31 11:31:34 11:31:37 11:31:39

8 9 10 11 12 13 14 15 16 17 18 19 20 21 22

MR. PICKETT:

Q.

If I demonstrate

statistically that something has a confidence level of 90 percent -A. Q. Yes. -- you would agree that it's more probably

true than not? MR. WILKES: THE WITNESS: MR. PICKETT: A. Same objections. I would not agree. Q. Why not?

Based on what you told me, there's no

reason for believing that. Q. Well, tell me why. That's not a reason.

That's just a repeat -- repetition of your opinion. A. I can always come up with a 90 percent

confidence interval if I have any data.

TEXT REMOVED - NOT RELEVANT TO MOTION

Merrill Legal Solutions [phone]

3c2f762b-a991-489c-93b9-1d08b41d0163

BRUCE SPENCER, PH.D. June 4, 2010 HIGHLY CONFIDENTIAL - ATTORNEYS' EYES ONLY
Page 99

TEXT REMOVED - NOT RELEVANT TO MOTION

11:33:13 11:33:15 11:33:16 11:33:18 11:33:23 11:33:27 11:33:32

11 12 13 14 15 16 17

Q.

What does a 90 percent confidence interval

mean to you, Dr. Spencer? A. A 90 percent confidence interval is a

random interval constructed so that with 90 percent probability, or for 90 percent of the possible samples, your interval will include the value that you're interested in.

TEXT REMOVED - NOT RELEVANT TO MOTION

Merrill Legal Solutions [phone]

3c2f762b-a991-489c-93b9-1d08b41d0163

BRUCE SPENCER, PH.D. June 4, 2010 HIGHLY CONFIDENTIAL - ATTORNEYS' EYES ONLY
TEXT REMOVED - NOT RELEVANT TO MOTION
12:12:29 12:12:31 12:12:34 12:12:38 12:12:42 12:12:48 12:12:51 12:12:54 12:12:57 12:13:03 12:13:09 12:13:09 12:13:11 12:13:13 12:13:19 12:13:27 12:13:30 12:13:33 12:13:35 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20

Page 116

Q.

Well, what does a 51 percent confidence

interval mean? A. Well, it means that you've constructed

it -- it means to me that you've constructed it in order that it have 51 percent probability of covering the population value under repeated sampling. That's what it means to me. But that you intend it to be so and that it is so are two different things. Q. A. Q. A. Q. So if you properly do the sampling -Right. -- and analyze it properly -Right. Right.

-- then you've constructed an analysis

that will, on repeated samples, have the value of interest within the bounds, within the bounds of the range 51 percent of the time -- at least 51 percent of the time. A. No. Right?

TEXT REMOVED - NOT RELEVANT TO MOTION

Merrill Legal Solutions [phone]

3c2f762b-a991-489c-93b9-1d08b41d0163

BRUCE SPENCER, PH.D. June 4, 2010 HIGHLY CONFIDENTIAL - ATTORNEYS' EYES ONLY
Page 121
12:29:15 12:29:28 12:29:30 12:29:34 12:29:40 12:29:43 12:29:47 12:29:49 12:29:50 12:29:52 12:29:57 12:30:12 12:30:15 12:30:18 12:30:24 12:30:33 12:30:37 12:30:40 12:30:45 12:30:50 12:30:53 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21

Q.

Let's go back to your report, please. I want to go to the sampling with

replacement versus sampling without replacement issue. Your position is that sampling without replacement is preferable to sampling with replacement? A. Q. Correct. But isn't it the case that sampling with

replacement is an acceptable method of sampling under sampling theory? A. It can be used. It's often not the best

thing to use, but it -- yes, it can be used. Q. And you've described it in your own Right? I sometimes use theory for

literature as an option. A. I don't know.

sampling with replacement when I have sampling with unequal probabilities. I'm not sure that I've used

it as a way of sampling populations where we sample them with replacement. recollecting that. That's -- I'm not

TEXT REMOVED - NOT RELEVANT TO MOTION

Merrill Legal Solutions [phone]

3c2f762b-a991-489c-93b9-1d08b41d0163

BRUCE SPENCER, PH.D. June 4, 2010 HIGHLY CONFIDENTIAL - ATTORNEYS' EYES ONLY
Page 123

TEXT REMOVED - NOT RELEVANT TO MOTION

12:33:34 12:33:35 12:33:37 12:33:40 12:33:43

16 17 18 19 20

Q.

And do you have any evidence that he was

lying when he so testified? MR. WILKES: THE WITNESS: Objection. Form.

I have no evidence that he

was lying, and I've never stated that he was lying.

TEXT REMOVED - NOT RELEVANT TO MOTION

Merrill Legal Solutions [phone]

3c2f762b-a991-489c-93b9-1d08b41d0163

BRUCE SPENCER, PH.D. June 4, 2010 HIGHLY CONFIDENTIAL - ATTORNEYS' EYES ONLY
Page 152

TEXT REMOVED - NOT RELEVANT TO MOTION

14:03:37 14:03:43 14:03:48 14:03:53 14:03:56 14:03:58 14:04:03 14:04:05

18 19 20 21 22 23 24 25

Q. test? A. Q.

Is your z statistic analysis a hypothesis

I haven't set it up that way. Well, isn't the hypothesis you're testing

that Dr. Levy did not perform a random sampling of the Critical Support population? A. I just said I didn't set it up as a

hypothesis test.

Merrill Legal Solutions [phone]

3c2f762b-a991-489c-93b9-1d08b41d0163

BRUCE SPENCER, PH.D. June 4, 2010 HIGHLY CONFIDENTIAL - ATTORNEYS' EYES ONLY
Page 153
14:04:06 14:04:09 14:04:11 14:04:12 14:04:14 14:04:16 14:04:18 14:04:19 14:04:19 1 2 3 4 5 6 7 8 9

Q.

Isn't that your hypothesis that you're

trying to test, whether you set it up that way or not? Isn't that what you did? MR. WILKES: THE WITNESS: Objection. Form.

I'm not doing a formal

hypothesis test in this analysis. MR. PICKETT: informal one? A. No. Q. Are you doing an

TEXT REMOVED - NOT RELEVANT TO MOTION

Merrill Legal Solutions [phone]

3c2f762b-a991-489c-93b9-1d08b41d0163

BRUCE SPENCER, PH.D. June 4, 2010 HIGHLY CONFIDENTIAL - ATTORNEYS' EYES ONLY
Page 199
TEXT REMOVED - NOT RELEVANT TO MOTION
15:14:30 15:14:33 15:14:36 15:14:40 15:14:43 15:14:46 15:14:49 15:14:53 15:14:58 15:14:59 15:15:02 15:15:03 15:15:04 15:15:06 15:15:07 15:15:09 15:15:12 15:15:15 15:15:21 15:15:24 15:15:31 15:15:33 15:15:39 15:15:41 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25

Q.

Does it impact the confidence interval for

the sampled measures? A. The underlying skewness, I believe, will

affect the coverage properties of the confidence intervals for the sampled measures, yes. Q. A. How? When you have large skewness, it appears

that the coverage will tend to be lower than the advertised rate. Q. samples? MR. WILKES: Objection. Form. Vague and Large skewness in the fully enumerated

ambiguous, asked and answered -MR. PICKETT: Q. What did you mean when

you said, when you tend to have -- I'm sorry, when you have large skewness? Did you mean large

skewness in the fully enumerated samples? A. I -- I meant large skewness with the

variable for which you're calculating a confidence interval, whether it's fully enumerated or sampled. Q. How does skewness in one measure affect

the confidence interval of another measure? A. Q. Directly, it does not. Well, how does it, then?

Merrill Legal Solutions [phone]

3c2f762b-a991-489c-93b9-1d08b41d0163

BRUCE SPENCER, PH.D. June 4, 2010 HIGHLY CONFIDENTIAL - ATTORNEYS' EYES ONLY
Page 200
15:15:46 15:15:48 15:15:50 15:15:53 15:15:55 15:15:56 15:15:58 15:16:07 15:16:08 15:16:15 15:16:18 15:16:21 15:16:24 15:16:27 15:16:29 15:16:30 15:16:32 15:16:36 15:16:40 15:16:45 15:16:49 15:16:52 15:16:59 15:17:01 15:17:02 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25

MR. WILKES: THE WITNESS: more fully? "it." MR. PICKETT:

Objection.

Form.

Can you state the question

I'm not understanding the reference of

Q.

Well, is it the case

that skewness in one measure does not affect the confidence interval of another measure? A. That's right. We calculate the confidence Finding that

interval one measure at a time.

skewness is a problem for some measures where we can test it gives us insight that it's also a problem for measures where we can't directly test the coverage of the confidence intervals. Q. So you're assuming some kind of

relationship between the samples? A. There's an established mathematical And

relationship between coverage and skewness. here, we had opportunity, we had data, we had

measures that Mandiant had fully enumerated the -we had some measures that were fully enumerated, so we could actually test empirically what the coverage was and was the skewness associated with problematic coverage. Q. So is it your testimony that there's an

established mathematical relationship between

Merrill Legal Solutions [phone]

3c2f762b-a991-489c-93b9-1d08b41d0163

BRUCE SPENCER, PH.D. June 4, 2010 HIGHLY CONFIDENTIAL - ATTORNEYS' EYES ONLY
Page 201
15:17:04 15:17:09 15:17:10 15:17:14 15:17:16 15:17:17 15:17:18 15:17:20 15:17:21 15:17:24 15:17:25 15:17:27 15:17:28 15:17:29 15:17:32 15:17:36 15:17:39 15:17:42 15:17:45 15:17:48 15:17:51 15:17:52 15:17:53 15:17:55 15:17:56 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25

coverage and skewness across measures? A. Q. That's not what I said. Well, I asked you in my question you're

responding to, I said, you're assuming some kind of relationship between the samples. So I'm asking you to look at a relationship between one sample and the other. you testified that there was an established mathematical relationships between coverage and skewness. Is that across measures or isolated to a single measure? MR. WILKES: Objection. Form. Compound, And

Argumentative, vague and ambiguous. asked and answered. THE WITNESS:

Misquotes the testimony. The skewness for one measure

is what directly affects the coverage rate of the confidence interval for that measure. And once you

take that into account, the skewness for another measure doesn't matter. MR. PICKETT: measures? MR. WILKES: answered. THE WITNESS: The skewness for one Objection. Form. Asked and Q. So it's not across the

Merrill Legal Solutions [phone]

3c2f762b-a991-489c-93b9-1d08b41d0163

BRUCE SPENCER, PH.D. June 4, 2010 HIGHLY CONFIDENTIAL - ATTORNEYS' EYES ONLY
Page 202
15:17:57 15:18:05 15:18:08 15:18:12 15:18:14 1 2 3 4 5

variable doesn't affect the coverage rate of the skewness for another variable, except to the extent that they're correlated, and that the skewness measures for the two variables tend to be related to each other.

TEXT REMOVED - NOT RELEVANT TO MOTION

Merrill Legal Solutions [phone]

3c2f762b-a991-489c-93b9-1d08b41d0163

BRUCE SPENCER, PH.D. June 4, 2010 HIGHLY CONFIDENTIAL - ATTORNEYS' EYES ONLY
Page 211

TEXT REMOVED - NOT RELEVANT TO MOTION

15:30:02 15:30:04 15:30:18 15:30:19 15:30:20 15:30:23 15:30:26 15:30:28 15:30:31 15:30:33 15:30:39 15:30:40 15:31:17 15:31:18 15:31:20

11 12 13 14 15 16 17 18 19 20 21 22 23 24 25

Q.

On the next page of your report, in

paragraph 6.34, at the very end of the page, you state: This point applies with particular force to his confidence intervals for measures available only for his sample, because their coverage rates could be much less than the 90 percent that Dr. Levy claims. A. Q. A. I see that statement. What did you mean by that statement? I'll have to read the paragraph. (Examining document.) So I say: Using the Student's t distribution does not solve the larger problem, which is that

Merrill Legal Solutions [phone]

3c2f762b-a991-489c-93b9-1d08b41d0163

BRUCE SPENCER, PH.D. June 4, 2010 HIGHLY CONFIDENTIAL - ATTORNEYS' EYES ONLY
Page 212
15:31:22 15:31:24 15:31:28 15:31:31 15:31:33 15:31:35 15:31:39 15:31:41 15:31:47 15:31:54 15:31:56 15:32:00 15:32:02 15:32:04 15:32:08 15:32:14 15:32:18 15:32:21 15:32:24 15:32:27 15:32:32 15:32:34 15:32:38 15:32:41 15:32:44 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25

for some measures, his confidence intervals can have coverage rates that are much poorer than what Dr. Levy claims. This point

applies with particular force to his confidence intervals for the measures available only for his sample, because their coverage rates could be much less than the 90 percent that Dr. Levy claims. Yeah. So I don't know why I said this

point applies with particular force, because these are the confidence intervals that we're most interested in. interested in. only measures. So -- but the point applies there that having seen that for the population measures, many of the coverage rates are less than 90 percent, and sometimes considerably less than 90 percent, the same may hold for the sampled measures as well, even though we can't directly verify that. Q. Right. May hold, but you don't know one Right? Objection. Form. Or these are the ones that we are The ones available for the sample

way or the other.

MR. WILKES: THE WITNESS:

Well, we have -- we have

some related evidence such as Exhibit 3100.

Merrill Legal Solutions [phone]

3c2f762b-a991-489c-93b9-1d08b41d0163

BRUCE SPENCER, PH.D. June 4, 2010 HIGHLY CONFIDENTIAL - ATTORNEYS' EYES ONLY
Page 213
15:32:51 15:32:52 15:32:54 15:32:57 15:32:59 15:33:01 15:33:02 15:33:03 15:33:07 15:33:09 15:33:11 15:33:11 15:33:15 15:33:17 15:33:19 15:33:21 15:33:25 15:33:28 15:33:30 15:33:34 15:33:35 15:33:36 15:33:38 15:33:39 15:33:40 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25

MR. PICKETT: A.

Q.

Well, what did --

Indicating that for some of them, the

coverage rates we can be confident will be less than 90 percent. Q. percent. When you said much less than 90 percent, what did you mean? A. Q. A. Q. In the 70s. And -Or lower. -- Exhibit 3100 doesn't indicate 70s. It Well, you said could be much less than 90

indicates 84 and above, doesn't it? MR. WILKES: THE WITNESS: Objection. Form.

It indicates that if the

sample skewness, as you report, is equal to the population skewness. But if the population

skewness is double what the sample skewness is, then your coverage is likely going to be much lower. MR. PICKETT: Q. But you have no evidence

that it's double do you? MR. WILKES: THE WITNESS: Objection. Form.

Well, we can see that for

some of the population measures, when you calculate

Merrill Legal Solutions [phone]

3c2f762b-a991-489c-93b9-1d08b41d0163

BRUCE SPENCER, PH.D. June 4, 2010 HIGHLY CONFIDENTIAL - ATTORNEYS' EYES ONLY
Page 214
15:33:43 15:33:46 15:33:50 1 2 3

the skewness for the sample, the skewness for the sample considerably underestimates the skewness in the population.

TEXT REMOVED - NOT RELEVANT TO MOTION

15:34:42 15:34:43 15:34:47 15:34:50 15:34:57 15:34:59 15:35:03

19 20 21 22 23 24 25

MR. PICKETT:

Q.

So what evidence do you

have that the coverage rates for the sampled measures would be 70 percent or lower? MR. WILKES: THE WITNESS: Objection. Form.

Well, we could -- and I have

to do this more carefully than I'm doing it right now.

Merrill Legal Solutions [phone]

3c2f762b-a991-489c-93b9-1d08b41d0163

BRUCE SPENCER, PH.D. June 4, 2010 HIGHLY CONFIDENTIAL - ATTORNEYS' EYES ONLY
Page 215
15:35:05 15:35:09 15:35:12 15:35:16 15:35:20 15:35:24 15:35:27 15:35:29 15:35:32 15:35:35 15:35:37 15:35:39 15:35:41 15:35:44 15:35:47 15:35:49 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16

Look at the comparison of the sample skewness measures and the population skewness measures for the fully enumerated variables, and see, do we have some significant -- or I shouldn't say significant -- some appreciable understatement of the population skewness. MR. PICKETT: A. Q. Have you done that? And for some of them,

I've looked at it.

there is a substantial underestimation by the sample skewness. Q. A. it out. Q. A. Which ones? We'd have to look at some tables and pick As I sit here, I don't know which ones. Is there an audit trail for your analysis? The tables exist. Any one of us could do

the analysis.

TEXT REMOVED - NOT RELEVANT TO MOTION

Merrill Legal Solutions [phone]

3c2f762b-a991-489c-93b9-1d08b41d0163

BRUCE SPENCER, PH.D. June 4, 2010 HIGHLY CONFIDENTIAL - ATTORNEYS' EYES ONLY
Page 223

TEXT REMOVED - NOT RELEVANT TO MOTION

15:59:18 15:59:21 15:59:27 15:59:29 15:59:36 15:59:41 15:59:47 15:59:51 15:59:52 15:59:55 15:59:58 15:59:59 16:00:00 16:00:02 16:00:06 16:00:16 16:00:17

9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25

Q. A.

In other words, you eyeball it? I wouldn't say eyeballing is. An example

would be with the correlation tables that we recently turned over, we have the RMS correlation size. And if the RMS correlation size is .6, then

the square of that is about .36, which means you'd get an effective reduction in sampling variance of about 35 percent. Q. And you did this analysis with respect to

the degree of the lower variance using stratification? A. Q. Yes. Did you leave an audit trail for us to

figure out what you did in that respect? MR. WILKES: THE WITNESS: Objection. Form.

Well, at the time I wrote

the report, I hadn't done that calculation.

Merrill Legal Solutions [phone]

3c2f762b-a991-489c-93b9-1d08b41d0163

BRUCE SPENCER, PH.D. June 4, 2010 HIGHLY CONFIDENTIAL - ATTORNEYS' EYES ONLY
Page 224
16:00:23 16:00:24 16:00:27 16:00:27 16:00:36 16:00:37 16:00:42 16:00:45 16:00:52 16:00:57 16:01:01 1 2 3 4 5 6 7 8 9 10 11

MR. PICKETT:

Q.

Well, did you leave an

audit trail after you did the calculation, whenever you did it? MR. WILKES: THE WITNESS: trail right now. Objection. Form.

Well, I can leave an audit

I'll just tell you how to do it.

That's you take the RMS correlation for one of the population measures and square it, and that gives you the approximate variance reduction if you do proportional allocation stratifying on that variable.

TEXT REMOVED - NOT RELEVANT TO MOTION

Merrill Legal Solutions [phone]

3c2f762b-a991-489c-93b9-1d08b41d0163