[00:00:08] Speaker 01:
23-1877, Marmon, Inc.

[00:00:11] Speaker 01:
vs. United States.

[00:00:48] Speaker 06:
Whenever you're ready, sir.

[00:00:50] Speaker 06:
Good morning.

[00:00:51] Speaker 06:
May it please the court.

[00:00:53] Speaker 06:
I will start with the minor correction to Mom and Inc's cost reconciliation.

[00:00:58] Speaker 06:
Commerce's refusal to accept a minor correction to the cost reconciliation is unsupported by substantial evidence.

[00:01:06] Speaker 06:
Now, Marmon is a Canadian company.

[00:01:09] Speaker 06:
Its financial statements are expressed in Canadian dollars, and it reported its production costs to commerce in Canadian dollars.

[00:01:17] Speaker 06:
So, in reconciling the company's audited cost of goods sold to its reported costs,

[00:01:22] Speaker 06:
It's necessary that all the adjustments in between must also be expressed in Canadian dollars.

[00:01:28] Speaker 06:
However, inadvertently in preparing its Excel spreadsheet reconciliation, Monmin failed to ensure that one value, one line was entirely expressed in Canadian dollars.

[00:01:39] Speaker 06:
So Monmin submitted a corrected cost reconciliation.

[00:01:43] Speaker 06:
together with support, and Commerce unreasonably rejected the correction.

[00:01:50] Speaker 06:
First, Commerce reasoned that the exchange rate correction that Marmon sought to make was already accounted for in the already cost of goods sold, the starting position for the reconciliation, and in the reported cost, the ending point.

[00:02:07] Speaker 06:
But this is totally illogical.

[00:02:10] Speaker 06:
of course, agree that the cost of goods sold and the reported costs were correctly expressed in Canadian dollars.

[00:02:16] Speaker 06:
But the point is, for the reconciliation to work, all the adjustments in between also need to be expressed in Canadian dollars.

[00:02:23] Speaker 06:
One of them wasn't.

[00:02:24] Speaker 06:
It was just one line of an Excel spreadsheet that Monman prepared, and there's absolutely no double counting.

[00:02:32] Speaker 05:
I realize this may seem like

[00:02:34] Speaker 05:
You know, the flea on the tail of the dog.

[00:02:37] Speaker 05:
But one of the points Commerce makes, at least here, is that you relied on a one-year average for the exchange rate that was incorrect or unsupported.

[00:02:57] Speaker 05:
Suppose we agreed with you on everything else.

[00:03:02] Speaker 05:
Why would that not be enough to reject this point?

[00:03:07] Speaker 05:
Commerce's finding with respect to the exchange rate is particularly unreasonable.

[00:03:11] Speaker 05:
I know you say they have these things on their website, but so.

[00:03:17] Speaker 06:
Well, what that means is that Momin reported and used in its correction the exchange rate that it used in the normal course of business during the period of investigation in its cost system as reported on the record.

[00:03:30] Speaker 05:
And a different exchange rate for each of the invoice sales or use the same exchange rate for the group during this is a six month period, right, in which you mistakenly used US dollars where you should have Canadian.

[00:03:46] Speaker 05:
Do you use a single exchange rate when you submitted the L1 correction, even though there were different sales during that period?

[00:04:01] Speaker 06:
Right.

[00:04:01] Speaker 06:
Well, in Marmon's accounting system, the issue was that in 2018, whereas its accounting system left US dollar purchases unconverted, in 2019, its accounting system would automatically convert US dollar purchases to Canadian dollars using a fixed rate of exchange.

[00:04:22] Speaker 05:
Fixed?

[00:04:22] Speaker 05:
I mean, fixed for what period?

[00:04:24] Speaker 06:
The date?

[00:04:25] Speaker 06:
For the entire year 2019.

[00:04:27] Speaker 06:
But in making this correction, this was in Marmon's accounting system.

[00:04:31] Speaker 06:
But in making the correction, what Marlin did was it used the actual exchange rate that it used in its cost reporting system.

[00:04:46] Speaker 06:
which is almost identical to the exchange rate that its accounting system would automatically make in 2019.

[00:04:54] Speaker 06:
Marmon used the same exchange rate for the 2018 purchases.

[00:04:58] Speaker 06:
And we're talking about an exchange rate to reject the correction simply because commerce says that an exchange rate

[00:05:05] Speaker 06:
is unsupported, is completely unreasonable.

[00:05:07] Speaker 02:
Can I just insert one more question?

[00:05:09] Speaker 02:
This is a smaller flee on the tail of the dog than Judge Toronto's was.

[00:05:14] Speaker 02:
But I know we can't talk about the numbers here.

[00:05:17] Speaker 02:
But the amici spend quite a bit of time arguing what rate this period is for.

[00:05:25] Speaker 02:
Can you tell us, is this for all of 2018?

[00:05:27] Speaker 02:
All of a period of interest?

[00:05:31] Speaker 02:
You're just the second half of 2018?

[00:05:34] Speaker 06:
The exchange rate that Norman used in the correction was for the entire year 2018.

[00:05:42] Speaker 06:
But nevertheless, Commerce, again, could look on its website and corroborate that it's an accurate exchange rate.

[00:05:49] Speaker 06:
The real issue isn't the exchange rate.

[00:05:51] Speaker 04:
It's just that one line of the spreadsheet of the reconciliation is- So are you saying that Commerce, instead of using the exchange rate you proposed, should have

[00:06:01] Speaker 04:
unilaterally gone to its website and figured out what the correct exchange rate should have been?

[00:06:07] Speaker 06:
No, no, I'm not arguing that.

[00:06:09] Speaker 06:
What I'm saying is that it's unreasonable to reject an exchange rate on the ground that it's unsupported when the administering agency can

[00:06:19] Speaker 06:
maintains daily exchange rates for anti-dumping calculation purposes and could easily check to make sure that it's comfortable that the exchange rate we used and reported is accurate.

[00:06:34] Speaker 02:
And there's some small disparity, difference in the numbers.

[00:06:38] Speaker 02:
Is that a rounding up issue?

[00:06:40] Speaker 02:
Is that just around me?

[00:06:42] Speaker 06:
A small difference in the exchange rates?

[00:06:44] Speaker 06:
Yeah.

[00:06:46] Speaker 06:
There's a small difference in the exchange rate that Marmon's system, accounting system, automatically used in 2019 to convert US dollar purchases to Canadian dollar purchases.

[00:06:58] Speaker 06:
the actual exchange rate that Marmon incurred during 2018.

[00:07:03] Speaker 06:
There is a small discrepancy there, but it's very small, and it pales in comparison to the discrepancy that the error created by virtue of the fact that Marmon inadvertently failed to ensure that one line in a cost reconciliation spreadsheet was entirely expressed in Canadian dollars.

[00:07:21] Speaker 06:
And accuracy should be commerce's primary concern.

[00:07:24] Speaker 05:
Yeah, but they depend on you for information.

[00:07:28] Speaker 05:
So and do I remember right?

[00:07:31] Speaker 05:
The period of investigation is basically mid 2018 to mid 2019, right?

[00:07:36] Speaker 05:
So and you changed your internal accounting system as of January 1st.

[00:07:42] Speaker 05:
So from the second half of that period, all the conversions were made uncontested apparently.

[00:07:50] Speaker 05:
The problem was that the conversions were not made for the first half of the period of investigation, which fell within calendar year 2018.

[00:07:57] Speaker 05:
And so you needed an exchange rate or multiple exchange rates for the various sales that were made during that period.

[00:08:04] Speaker 05:
Sales, I'm sorry, this is purchases by you from your subsidiary or affiliate.

[00:08:10] Speaker 06:
Correct.

[00:08:10] Speaker 06:
We're talking about one line in the spreadsheet.

[00:08:13] Speaker 06:
that we're not going to be able to do that.

[00:08:18] Speaker 06:
So that's why we're not going to be able to do that.

[00:08:23] Speaker 06:
So that's why we're not going to be able to do that.

[00:08:27] Speaker 06:
So that's why we're not going to be able to do

[00:08:30] Speaker 06:
You just took the number from its accounting system, which was the purchase value for those six months, which was in US dollars.

[00:08:36] Speaker 05:
Right.

[00:08:36] Speaker 05:
Now we're just talking about this seemingly small, but maybe consequential issue at the end, that when you said convert by one point whatever it is, that they said, why should we use that number?

[00:08:55] Speaker 05:
What's your justification for using that number?

[00:08:59] Speaker 05:
They now at least here say that unjustified choice alone was enough for them to reject the L1 change.

[00:09:09] Speaker 06:
Right, Your Honor, that is commerce's explanation.

[00:09:12] Speaker 06:
And our argument is that that was completely unreasonable, because we're not talking about an error in reported costs.

[00:09:22] Speaker 06:
We're talking about an error in an exchange rate that can easily be validated by commerce itself.

[00:09:28] Speaker 06:
And in addition to the exchange rate, Marmon submitted, for the record, a schedule of all the invoices at issue, the US dollar and all of which

[00:09:38] Speaker 06:
That schedule clearly showed that the purchases in 2018 were not properly converted to Canadian dollars.

[00:09:47] Speaker 06:
They were left in US dollars.

[00:09:49] Speaker 06:
We explained that schedule of inverses at page 48 of Marmon's appellants brief.

[00:09:55] Speaker 06:
And the appellees, it's unassailable what it demonstrates, and the appellees failed to respond.

[00:10:01] Speaker 06:
Your Honors, if I may, I think I need to move on to the Cohen's D argument.

[00:10:07] Speaker 06:
Commerce's reliance on the Cohen's Deed coefficient and its consequent calculation of Marlin's dumping margin using the average to transaction calculation method cannot be sustained.

[00:10:20] Speaker 02:
In step, this court recognized... Can I ask, just because time is short, so I don't want to... Yeah.

[00:10:25] Speaker 02:
If we agree with you, what's the next step?

[00:10:28] Speaker 02:
And I'll ask this of the government as well.

[00:10:31] Speaker 02:
I mean, is there a way... Is everybody assured there's another way to do this?

[00:10:36] Speaker 02:
What do you envision the next step?

[00:10:41] Speaker 02:
If we were to say the same thing we said in Stubb and the same thing we've said before and not satisfied that commerce has given us a sufficient basis for being okay with that, then what happens?

[00:10:55] Speaker 06:
There should be no next step for commerce.

[00:10:58] Speaker 06:
There should be no remand for commerce to correct this.

[00:11:01] Speaker 06:
Commerce has been given a chance, and commerce is not using the Cohen's D coefficient accurately.

[00:11:08] Speaker 06:
Therefore, commerce's decision that there was a pattern of significant price differences cannot be sustained.

[00:11:15] Speaker 06:
It's not reasonable.

[00:11:16] Speaker 06:
It cannot be sustained.

[00:11:17] Speaker 06:
Commerce should not be given a second chance.

[00:11:20] Speaker 06:
It should recalculate Marmon's dumping margin using the standard average to average

[00:11:24] Speaker 06:
transaction comparison method.

[00:11:27] Speaker 05:
Can I ask a... I don't remember this, I should, but when Commerce did Cohen's D here, did it, in the denominator, use a weighted average of the two different variances or standard deviations?

[00:11:44] Speaker 05:
Doesn't matter.

[00:11:44] Speaker 06:
Commerce used a... its normal practice is to use a simple average of the standard deviations for the two different datasets.

[00:11:52] Speaker 05:
OK, but you haven't separately challenged that.

[00:11:54] Speaker 05:
That was the issue in Mid-Continent.

[00:11:56] Speaker 05:
Correct.

[00:11:56] Speaker 05:
But it also, but it recognized that there were two, that the standard deviations were distinct between the control group and the test group.

[00:12:05] Speaker 06:
Right, right.

[00:12:06] Speaker 05:
That's the Mid-Continent case.

[00:12:07] Speaker 05:
And then it averaged.

[00:12:09] Speaker 06:
Our main argument here is that commerce is probably.

[00:12:12] Speaker 05:
Do you happen to know where that is, where they said we're doing sigma prime and not sigma?

[00:12:17] Speaker 06:
Sure, in commerce is both its preliminary decision memo and its final decision.

[00:12:24] Speaker 05:
Only JA numbers count in answer to a question.

[00:12:27] Speaker 06:
I don't have it at the moment, your honor.

[00:12:30] Speaker 06:
I apologize.

[00:12:31] Speaker 06:
Our main argument is that commerce's excuse or justification for not ensuring that the assumptions underlying the Cohen's decoefficient are satisfied is wrong.

[00:12:45] Speaker 06:
Commerce's claim is that when it uses the Cohen's decoefficient,

[00:12:53] Speaker 06:
comparing data sets that consist of complete populations, complete populations of the US sales.

[00:12:59] Speaker 06:
And Commerce claims that under this circumstance, when you compare the data sets of populations as opposed to samples, the Cohen's d assumptions no longer apply.

[00:13:09] Speaker 06:
But this is completely wrong.

[00:13:11] Speaker 06:
Professor Cohen himself, in explaining the importance of the assumptions,

[00:13:16] Speaker 06:
did so in the context of comparing populations.

[00:13:20] Speaker 06:
And you can see this at appendix 4736.

[00:13:23] Speaker 06:
In addition, Professor Cohen calculated the small, medium, and large thresholds in relation all in the context of comparing data sets that are populations, and also while ensuring that the three assumptions are satisfied.

[00:13:40] Speaker 06:
And you can see this at appendix 4739 through page 4742.

[00:13:46] Speaker 06:
The Cohen's decoefficient itself that Professor Cohen calculated and his thresholds for small, medium, and large are tied to percentages of non-overlap.

[00:13:59] Speaker 06:
And it's these percentages of non-overlap that give the Cohen's decoefficient and the small, medium, and large thresholds meaning in the first place.

[00:14:09] Speaker 06:
However, if any of the assumptions, normal distribution,

[00:14:13] Speaker 06:
equivalent variances, equal size.

[00:14:16] Speaker 06:
If any of those assumptions is violated, then the Cohen's de-coefficient will not correspond to the same percentages of non-overlap that relate to the small, medium, and large cutoffs.

[00:14:29] Speaker 06:
And then the de-coefficient is essentially meaningless.

[00:14:32] Speaker 04:
Do you happen to know the data in this case, how close the normal distribution

[00:14:39] Speaker 04:
the various curves are or whether the variance is more or less the same, maybe not identical, and then also whether the size of the various populations is kind of equal.

[00:14:54] Speaker 06:
Yes, Your Honor.

[00:14:55] Speaker 06:
In fact, during the remand proceeding, we submitted a demonstration that the three assumptions, and in particular the assumptions of normal distributions and equivalent variances, were not satisfied.

[00:15:08] Speaker 06:
We did say at appendix 3987 through 4663, but Commerce rejected that submission.

[00:15:15] Speaker 04:
They say it was- There's something in Cohen itself that suggests

[00:15:25] Speaker 04:
So therefore, that's where you go to sigma prime.

[00:15:32] Speaker 06:
Your Honor, I believe you're referring to, and this is something that the government pointed out, equation 2.3.

[00:15:39] Speaker 06:
point 3.2.

[00:15:40] Speaker 06:
But there, and you're correct in that scenario, Professor Cohen did say that the requirement of equivalent standard deviations could be relaxed somewhat.

[00:15:51] Speaker 06:
But he also articulated that only moderate departures from the three assumptions could be tolerated.

[00:15:57] Speaker 06:
And he stated that he indicated, it's at appendix 4734 through 4735, that if- I guess what I'm trying to get at, as I think you can tell based on your

[00:16:09] Speaker 04:
response is that maybe you don't need the assumptions to be perfectly satisfied, that maybe it's now an in the weeds factual dispute over what degree of dissimilarity you can have with respect to normal distribution, equal variance, and size.

[00:16:32] Speaker 06:
Your Honor, in the one scenario that you were referring to, equation 2.3.2, again, Professor Cullen did indicate that there could be a modest departure, but the overall point is the same, which is that the assumptions do matter.

[00:16:48] Speaker 06:
They are important.

[00:16:49] Speaker 06:
They cannot be completely disregarded.

[00:16:51] Speaker 05:
In fact, in that very example... Let me just say, I guess this is a different, I don't know,

[00:16:58] Speaker 05:
formulation of what I think is the same question.

[00:17:02] Speaker 05:
With the rejection of your submission about your data, which I do not see was appealed, how have you established prejudice from what I will assume for purposes of this question was an erroneous reliance on the point date figure of Cohen's d.

[00:17:26] Speaker 06:
Well, if I understand the question, Your Honor, I mean, our argument is... It relates to what Judge Chen said, right?

[00:17:33] Speaker 05:
I mean, sometimes it might be, you know, a small, I want to say variation, I don't mean variance, small variation from the assumptions of same variance between the two groups won't make a big difference in deciding whether the .8

[00:17:54] Speaker 05:
threshold figure, which is, I think, what all of this is about, has practical significance for identifying a real-world practical significant difference between the prices in the two groups, which is what we're talking about.

[00:18:07] Speaker 05:
But if you haven't established why your numbers would produce a different analysis, how have you shown prejudice from the erroneous assumption reliance on the Cohen-Steve figure?

[00:18:23] Speaker 06:
Well, Your Honor, we made a submission that was unreasonably rejected.

[00:18:27] Speaker 06:
But beyond that, it's really Commerce's responsibility to ensure that if it's going to rely on a Cohen's D coefficient to find a pattern of significant price differences, it has to use the Cohen's D coefficient correctly.

[00:18:42] Speaker 06:
So Commerce has the data.

[00:18:44] Speaker 06:
It has the US sales data.

[00:18:46] Speaker 06:
It can test whether the three assumptions are satisfied.

[00:18:49] Speaker 06:
And commerce's failure to do so here is impermissible, and it cannot be sustained.

[00:18:55] Speaker 06:
Your Honors, we are talking about price differences of less than 1%.

[00:19:00] Speaker 06:
And commerce claimed, based on blind reliance and the Cohen's decoefficient, that those price differences are large.

[00:19:07] Speaker 06:
Equivalent in Professor Cohen's words,

[00:19:09] Speaker 06:
to the difference in IQ between a college graduate and a student with a 50-50 chance of passing high school.

[00:19:16] Speaker 06:
Objectively, no reasonable person would agree that price differences of less than 1% are on par with

[00:19:24] Speaker 06:
That difference in IQ between a college graduate and a student who has only a 50-50 chance of passing high school.

[00:19:31] Speaker 06:
So your honor, it's commerce's error here.

[00:19:34] Speaker 06:
Commerce cannot rely on a Cohen's D coefficient without ensuring that it's using it correctly.

[00:19:41] Speaker 06:
And the way commerce uses the Cohen's D coefficient without adherence to any of the assumptions that make it meaningful is completely impermissible.

[00:19:55] Speaker 06:
No, no, no, I don't mean go away.

[00:19:57] Speaker 06:
I think that's all my time now unless you have other questions.

[00:20:04] Speaker 02:
All right.

[00:20:07] Speaker 02:
We'll hear from the government.

[00:20:08] Speaker 02:
Thank you.

[00:20:30] Speaker 03:
Good afternoon, Your Honors, and may it please the Court, I will pick up where my colleague left off in the discussion of the Cohen's D issue.

[00:20:42] Speaker 03:
I think the first important point to make here is that in the multiple cases that have addressed this issue,

[00:20:51] Speaker 03:
The Cohen's D coefficient is only one part of commerce's differential pricing analysis.

[00:20:59] Speaker 05:
But you have not even come close to establishing that if this first step, which is a filtering step, is unsupported, the rest makes up for it.

[00:21:09] Speaker 05:
I just don't see how commerce has established anything like that.

[00:21:16] Speaker 05:
The first step of .8, figuring out the D and segregating by .8 bigger to less and then going on to the next steps which have a .33 and all that stuff.

[00:21:27] Speaker 05:
I don't see how you commerce has established that even if we're completely wrong in using the .8 to do the filtering, the rest will cure all the problems of making this into a real world practical test.

[00:21:43] Speaker 03:
Well, I think, Your Honor, the issue there is that the appellants have tried to say that the Cohen's D coefficient is what commerce is using to establish the pattern that the statutory language calls for.

[00:22:01] Speaker 03:
And the point that I'm making is it's only one piece of that, right?

[00:22:06] Speaker 03:
That's true.

[00:22:07] Speaker 03:
So the issues that have been brought up here about, you know, the, you know, some hypotheticals that have been raised about what happens if, you know, the variances are different and how does that affect, you know, the ultimate coefficient and where does that go.

[00:22:24] Speaker 03:
Again, you have to run the process through to the end completing those other steps of the differential pricing analysis before

[00:22:35] Speaker 03:
would make the determination that they're meaningful.

[00:22:39] Speaker 03:
For example, it was raised towards the end of the appellant's briefing, although the hypothetical came from the Stupp case, had to do with the situation where what happens if you have a test and control group that have very small variances, right?

[00:22:55] Speaker 03:
And once you step through the math, what you wind up with is a coefficient, a Cohen's D coefficient, that's very, very high, right?

[00:23:05] Speaker 05:
So the dollar differences are trivial.

[00:23:07] Speaker 03:
Exactly, right.

[00:23:09] Speaker 03:
So OK, even if you have that, you still have to walk through the rest of the process.

[00:23:14] Speaker 03:
OK, in that situation, you're going to have most of the sales, if not all, pass the Co and Z test.

[00:23:22] Speaker 03:
We look at the ratio test, right, which has already been affirmed by this court as acceptable.

[00:23:27] Speaker 03:
Does it pass that 33% threshold or 66% threshold?

[00:23:32] Speaker 03:
assuming that it does, right?

[00:23:34] Speaker 03:
Because again, you walk through the math, you've got a lot of these passes.

[00:23:37] Speaker 03:
You still have that meaningful difference test.

[00:23:40] Speaker 03:
So commerce is still going to take a step back and look at these two groups and say, okay, is this difference meaningful?

[00:23:50] Speaker 03:
Is the A to A method de minimis and A to T is not?

[00:23:54] Speaker 03:
Or if neither method is de minimis, is there bigger than a 25% difference?

[00:23:59] Speaker 03:
You still have to go through that step.

[00:24:01] Speaker 05:
And what's the standard for determining whether the difference discovered is meaningful?

[00:24:13] Speaker 03:
In that final step of the differential pricing analysis, the meaningful difference

[00:24:17] Speaker 03:
Commerce will actually run both the A to A test on the sales and also run the A to T test on the sales.

[00:24:28] Speaker 05:
And the standard is... That's not a meaningful difference between the test and control group.

[00:24:34] Speaker 05:
That's just a meaningful difference between the two methodologies that they're choosing from.

[00:24:42] Speaker 03:
Right, but that's how they're determining whether or not a meaningful difference exists.

[00:24:47] Speaker 05:
They're looking at running the methodology on one and looking at what... If we can find some kind of violation, some kind of dumping with one method and not the other, and those methods turn out differently, non-trivially, then we'll do it.

[00:25:06] Speaker 05:
But that doesn't tell you...

[00:25:07] Speaker 05:
whether in, for example, the stub hypothetical about trivial dollar differences.

[00:25:16] Speaker 03:
Well, this kind of brings us back, I think, to the Cohen's D coefficient, right?

[00:25:21] Speaker 03:
Because all that the Cohen's D coefficient is attempting to do is tell us

[00:25:28] Speaker 03:
mathematically, are these two groups different?

[00:25:32] Speaker 03:
Is the difference between these two groups practically significant, is the language that Professor Cohen uses.

[00:25:39] Speaker 03:
So the language of the statute says that, you know, commerce will use this A to A method unless it can't account for, you know, these instances of what's been termed mask dumping.

[00:25:52] Speaker 03:
So they need a process, a methodology to determine whether or not when they're looking at these different regions, time periods, whatever,

[00:26:04] Speaker 03:
whether or not any differences that they see in what becomes the test and control group are, in fact, significant.

[00:26:13] Speaker 03:
And that's what they're using the Cohen's D coefficient to do.

[00:26:17] Speaker 05:
So let me, I guess, describe two different things, one of which I think

[00:26:24] Speaker 05:
Commerce did say, one of which perhaps Commerce could say.

[00:26:28] Speaker 05:
The thing that I think Commerce did say is, we're going to use the Cohen's d to tell us at the threshold whether there is a practically significant difference in the relevant outcome between the two groups.

[00:26:43] Speaker 05:
Even though Cohen, actually, because Cohen was about populations and not samples, which I think is just wrong.

[00:26:54] Speaker 05:
That's one thing.

[00:26:56] Speaker 05:
Another thing you could say is we recognize that Cohen said that he's getting the 0.8 as a threshold from consideration of normal distributions with, you know, maybe say equal variances or

[00:27:16] Speaker 05:
roughly equal variances.

[00:27:18] Speaker 05:
And importantly, same numbers, a few minutes in the two groups, unless you do weighted averaging, which is not at issue here, but was in mid-continent, and Commerce also did something unjustified, we thought, on that.

[00:27:38] Speaker 05:
But if Commerce said, we recognize the limits of what Cohen did for practical significance,

[00:27:47] Speaker 05:
He was considering one kind of distribution.

[00:27:49] Speaker 05:
But the general idea, and now we're going to do our own statistical analysis, methodological statistical explanation for why different distributions can be treated somewhat alike.

[00:28:02] Speaker 05:
And I don't see that second explanation.

[00:28:05] Speaker 05:
at all.

[00:28:06] Speaker 05:
All I see is we rely on Cohen.

[00:28:09] Speaker 05:
Cohen's greatly authoritative.

[00:28:12] Speaker 05:
And so we don't have to think anymore.

[00:28:15] Speaker 05:
And those are two very different things.

[00:28:17] Speaker 05:
That's what's, I guess, troubling me.

[00:28:20] Speaker 03:
I understand, Your Honor.

[00:28:21] Speaker 03:
I think that what Commerce has said here is they are

[00:28:32] Speaker 03:
I take your point about the explanation and how they phrased it.

[00:28:37] Speaker 05:
I think that what's- This is not a one-time thing, right?

[00:28:39] Speaker 05:
I mean, this is an ongoing methodological choice and seemingly a maybe even carbon copy defense of what they're doing.

[00:28:51] Speaker 03:
To a certain extent, yes, I would agree with that.

[00:28:54] Speaker 03:
I think that effectively, it's

[00:28:58] Speaker 03:
Commerce is using the Cohen's D coefficient for a particular purpose.

[00:29:06] Speaker 03:
They're taking this part of Cohen's analysis and saying that the equation

[00:29:16] Speaker 03:
will produce a coefficient that will show this, right?

[00:29:21] Speaker 03:
Now, the objection to that that's been raised is that, well, Cohen made all these assumptions, right?

[00:29:28] Speaker 03:
These equal variances and everything else that comes in there.

[00:29:32] Speaker 03:
And again, the response is, and I think this is kind of where, you know, maybe in artful in some way,

[00:29:41] Speaker 03:
But what commerce is essentially saying is we're using Cohen as the way that we're making this mathematical determination.

[00:29:51] Speaker 03:
So we're applying Cohen in this specific way.

[00:29:55] Speaker 03:
They're not using Cohen to conduct a comparison of sample groups.

[00:30:01] Speaker 03:
They're using it to compare entire populations of sales data, which they have.

[00:30:07] Speaker 04:
What in Cohen or any other literature

[00:30:13] Speaker 04:
tells us that when it comes to using whole populations as opposed to samples, you can throw away the assumptions.

[00:30:24] Speaker 04:
That seems to be the message and theme of commerce's position.

[00:30:30] Speaker 04:
And I looked at the sites in your brief.

[00:30:35] Speaker 04:
I looked at whatever sites commerce's redetermination referenced

[00:30:41] Speaker 04:
And I didn't see anything that remotely suggests that only when it comes to sampling do you need these assumptions to be in effect.

[00:30:53] Speaker 04:
But hey, it's important to understand that if you've got all the population data, you're good to go.

[00:30:59] Speaker 04:
Throw away the assumptions.

[00:31:01] Speaker 04:
Where is that in Cohen or in any other publication?

[00:31:08] Speaker 04:
You know, expert testimony.

[00:31:09] Speaker 04:
But I didn't see it.

[00:31:11] Speaker 03:
I can't pull a precise pull quote that says, you may throw away the assumptions if, right?

[00:31:18] Speaker 04:
But what commerce is- The best evidence they have for their personal theory on how to think about when you can and cannot comply with the assumptions.

[00:31:31] Speaker 03:
Two things jump to mind.

[00:31:32] Speaker 03:
First and foremost is Cohen's discussion itself, right?

[00:31:37] Speaker 03:
Where?

[00:31:39] Speaker 03:
It's in the appendix, the discussion in his paper, sections 2.2 to 2.3, where he's walking through all of this.

[00:31:50] Speaker 03:
The reason he's going through all of these discussions about the assumptions and the sampling

[00:31:56] Speaker 03:
there's this discussion of the t-test and all these different factors, is because he's trying to ensure that the samples being compared are reliable representations of the populations that they're intended to represent.

[00:32:16] Speaker 04:
And there's also... Where can I find a passage that tells me this is all sample sensitive that we need these assumptions?

[00:32:24] Speaker 04:
I know we can sit here and talk in generalities about what different competing and takeaway interpretations of Cohen might be, but where is there text I can look at that can give me some reasonable sense that commerce was onto something when it said, whole population data, forget about the assumptions?

[00:32:47] Speaker 03:
Well, there's a quote from Dr. Ellis' paper, if the quote will bear with me.

[00:32:56] Speaker 03:
Of course, I didn't put down the appendix citation.

[00:33:00] Speaker 03:
But we reference Dr. Ellis' paper talking about Dr. Cohen's analysis, where he says that, of course, the idea when making these comparisons would be if you had the entire population.

[00:33:16] Speaker 03:
But even in the examples that turn up in Cohen and have been mentioned here today, he's talking about the heights of young women and girls or the IQs within a population.

[00:33:28] Speaker 03:
They're referencing a population, OK, even if it's of this country, that's millions, tens of millions of people.

[00:33:37] Speaker 05:
But let me just see if I, I mean, this is, I guess, how I've been thinking about it.

[00:33:43] Speaker 05:
Dr. Cohen, he's a psychologist.

[00:33:48] Speaker 05:
You can't basically take measurements, IQ if that's what it was, on every 13 and 19 year old girl in the country.

[00:33:59] Speaker 05:
most psychological experiments, the groups that you are going to use in doing a comparison will be samples.

[00:34:09] Speaker 05:
Ideally, he says, of course, we'd have full populations.

[00:34:14] Speaker 05:
But the point of his whole Collins D is once you have groups, we're going to try to figure out if those groups

[00:34:25] Speaker 05:
differ significantly.

[00:34:27] Speaker 05:
It doesn't matter where you came, where you got the groups from.

[00:34:31] Speaker 05:
It does matter whether when you decide there's a significant difference between the groups, which is what Cohen's theory is about, whether you draw an inference about a whole population.

[00:34:44] Speaker 05:
But that's a later question.

[00:34:46] Speaker 05:
Though D-thing is about whether there is a difference, I call it effect size, the effect of being in group one versus group two, no matter how those groups came about.

[00:34:57] Speaker 03:
Right.

[00:34:58] Speaker 03:
And I think the first part of your statement is actually the answer to Judge Chen's question, which is that he's going through all of this text, all of this discussion about these different assumptions to ensure that his samples are accurate, are reliable.

[00:35:15] Speaker 03:
But when you have a population, reliability is not an issue.

[00:35:19] Speaker 03:
We're not afraid that the sample group is going to misrepresent the population because we have the entire population.

[00:35:26] Speaker 03:
And I think the reason why you don't have that statement, that clean pull quote, to say you can toss this out is that in most of the situations where they're making these comparisons, it's totally impractical to pull an entire population.

[00:35:41] Speaker 03:
But that's not the case here, nor does that make commerce's use of the equation and the resulting thresholds invalid.

[00:35:51] Speaker 04:
But you would agree with me that there's nothing in Cohen or any other literature that says these assumptions are needed because we're sampling.

[00:36:05] Speaker 04:
I mean, again, I didn't.

[00:36:07] Speaker 04:
There isn't.

[00:36:09] Speaker 03:
I think Colin does say that they're needed because they're sampling.

[00:36:14] Speaker 03:
That's his entire discussion about all of those assumptions.

[00:36:17] Speaker 04:
That's what I'm looking for.

[00:36:18] Speaker 04:
If there was something in Colin that said these assumptions are important because

[00:36:25] Speaker 04:
We're sampling.

[00:36:26] Speaker 04:
We don't have the full population.

[00:36:28] Speaker 03:
I apologize.

[00:36:29] Speaker 03:
I don't have the appendix site in here.

[00:36:31] Speaker 03:
The pages in Cohen's paper, which I'm confident are in the record, it runs after the introduction, which is the first few pages.

[00:36:39] Speaker 03:
It runs from about page seven through, I think, about page 40, which is where he starts to get into the various equations and so forth that we've been talking about.

[00:36:52] Speaker 03:
But again, when he starts with his

[00:36:55] Speaker 03:
his base equation and then goes into all of these other discussions, the entire point of those discussions is to ensure the reliability of the sample, right?

[00:37:07] Speaker 03:
And so, again, that's why, as this has come up in all of these cases, commerce is saying that that reliability is not an issue.

[00:37:18] Speaker 03:
I see my time has expired if the court has no further questions.

[00:37:23] Speaker 05:
Did you happen to remember what, I have the Alice thing, which is at 4805 in the appendix.

[00:37:31] Speaker 05:
What was it in, you referred to Alice earlier?

[00:37:33] Speaker 03:
Yes, so we noted in our briefing, we have a citation to... Page 64 of your brief?

[00:37:41] Speaker 03:
Yes.

[00:37:47] Speaker 05:
because you often can't get full population.

[00:37:50] Speaker 05:
So what?

[00:37:52] Speaker 05:
It's so what that's the point.

[00:37:54] Speaker 03:
I think it kind of goes to the premise of your earlier question, Your Honor, which is that once we have the samples, that Cohen is doing all of this work to get to a place where he has samples that are reliable representations of their populations.

[00:38:14] Speaker 05:
But he's not doing the comparison measure based on how the groups came about.

[00:38:24] Speaker 03:
I'm sorry, I'm not, I don't think I'm following.

[00:38:27] Speaker 05:
Whether you have the entire population of teenage girls in the country and you're actually measuring them and you're doing by age, the 13s and the 19s or whatever, the numbers were something, ages, right?

[00:38:41] Speaker 05:
Or whether you have a sample of 10,000

[00:38:45] Speaker 05:
of them, and you're doing 13 versus 19.

[00:38:50] Speaker 05:
The sampling process will be meaningful.

[00:38:55] Speaker 05:
That is a process you went through in sampling and what the distribution is like of the relevant variable that you're trying to measure will be important

[00:39:07] Speaker 05:
for whether, at the very end of the day, what you discover about the difference between the 13, 8-year-old group and the 19-year-old group, whether that difference, which is you have just two groups, the two cases you have out of a population of 10,000, another out of a population of, I don't know, 30 million or something, and the entire population, you're figuring out

[00:39:34] Speaker 05:
The difference in those two groups, when it's the 10,000, you want to draw an inference about the whole population.

[00:39:44] Speaker 05:
And it matters what the distribution is for whether the sample was good enough.

[00:39:50] Speaker 05:
The comparison between the two groups doesn't care one whit.

[00:39:55] Speaker 05:
about once you have the two groups, either out of the 10,000 or out of the 30 million, you're just trying to figure out if those two groups look practically significant in the relevant variable.

[00:40:11] Speaker 03:
Yes, that's correct.

[00:40:13] Speaker 05:
So that has nothing to do with sampling versus total population?

[00:40:16] Speaker 03:
That's correct.

[00:40:17] Speaker 05:
OK.

[00:40:21] Speaker 03:
Thank you.

[00:40:40] Speaker 00:
I'm not sure that I have all that much to add at this juncture.

[00:40:50] Speaker 00:
But one thing that I think might have gotten a little bit lost in all of this is why is Congress even doing this?

[00:40:56] Speaker 00:
Why do we care?

[00:40:57] Speaker 00:
And it's because there is this statutory

[00:41:00] Speaker 00:
requirement, or at least Congress has said, Congress, you can use, you may use this alternative calculation methodology if the following conditions are met.

[00:41:10] Speaker 00:
So Commerce reasonably has said, well, let's figure out some relatively predictable way of figuring out whether the conditions are met.

[00:41:18] Speaker 00:
And it really is as Commerce has adapted it, and I do not doubt, Commerce is not using Cohen's D in

[00:41:25] Speaker 00:
the sense that Cohen developed it for.

[00:41:27] Speaker 00:
He's not, Commerce is not a social scientist trying to design experiments and figure out in advance what sample size or population size would I need to get this kind of effect size.

[00:41:37] Speaker 00:
They're not doing that, but they've said we have a completely different kind of

[00:41:42] Speaker 00:
thing we need to do, it's got some analogous concepts to it, and we want to refine our old nails test, which basically used standard deviations and averages to make this kind of assessment of whether a response to sale prices differed across time, customer, region.

[00:42:03] Speaker 00:
To me, the Cohen's D is just the same kind of test.

[00:42:07] Speaker 00:
that commerce already had.

[00:42:09] Speaker 05:
It's perhaps refinement, but that general... But it has this very precise 0.8 figure that commerce... Well, is it very precise?

[00:42:18] Speaker 00:
Come on.

[00:42:19] Speaker 00:
It's an extremely specific number.

[00:42:21] Speaker 00:
Yes, but so is 25%.

[00:42:24] Speaker 00:
So is 33%.

[00:42:25] Speaker 00:
That's a specific number.

[00:42:27] Speaker 04:
But what if we disagree with commerce's point of view or find it unreasonable to say that

[00:42:35] Speaker 04:
Cohen's D formula, you don't need the assumptions when you have whole population data.

[00:42:43] Speaker 00:
I think even if you think, well, commerce, you're just not using Cohen's D the way Cohen meant it.

[00:42:49] Speaker 04:
Well, the way they want it to be understood.

[00:42:51] Speaker 00:
Well, you're not using it in the context Cohen meant it to be used, and you're not using it under the assumptions that Cohen describes with the use that he described.

[00:43:01] Speaker 00:
This is the sort of general method of analyzing price differences using standard deviations and averages is not somehow unique to Cohen's D. The Niels test had a somewhat similar use of averages and standard deviations.

[00:43:17] Speaker 00:
It just didn't trace itself to particular statistical literature.

[00:43:20] Speaker 00:
And in this way, I think you can understand commerce's adaptation and use of Cohen's D

[00:43:27] Speaker 00:
As similar to the rules of thumb, this court has already approved that commerce use for the ratio test or the meaningful difference test.

[00:43:34] Speaker 00:
Why did commerce choose 25% as the difference?

[00:43:37] Speaker 00:
It could have said 50, 33, 10, anything.

[00:43:40] Speaker 04:
Well, everything you're saying is maybe something commerce could say on remand.

[00:43:45] Speaker 04:
Because this is not the words that they used this time around.

[00:43:50] Speaker 00:
To Judge Prost's point then, what's next?

[00:43:52] Speaker 00:
I don't think what's next, as the appellant would have it, is just for this court to say, you know, three strikes and you're out commerce, you don't get to make these types of decisions.

[00:44:02] Speaker 00:
Recalculate Marmon's margin the way Marmon wants it.

[00:44:05] Speaker 00:
Instead, it's go back.

[00:44:06] Speaker 00:
Maybe they have to go back to the drawing board, but that's what would happen.

[00:44:09] Speaker 00:
It's go back to the drawing board.

[00:44:11] Speaker 00:
Maybe readapt the nails test.

[00:44:12] Speaker 00:
Maybe explain better.

[00:44:14] Speaker 00:
Maybe do what you said, Judge Toronto, and say,

[00:44:17] Speaker 00:
OK, we're not really doing Coen's D, but here's why our use of this similar or identical.

[00:44:22] Speaker 05:
The statistical literature that's been presented to us, some of it in the record, some of it not, acknowledge that, of course, when you're trying to figure out a effect size, there are a bunch of different statistical

[00:44:42] Speaker 05:
I don't mean tools, but when you're doing a statistical analysis, there's no single test.

[00:44:50] Speaker 05:
I don't mean that you go to the grocery store and there's a choice.

[00:44:53] Speaker 05:
This is all math.

[00:44:56] Speaker 05:
And the fact that Cohen came up with a particular way that's really quite remarkably simple and has certain virtues doesn't mean that there aren't variations on it that could be applied, but they have to be established.

[00:45:10] Speaker 05:
I think some of the articles use a much more sophisticated form of effect size measurement.

[00:45:22] Speaker 05:
But Commerz hasn't done the work to do that, to say, with prices, we don't know anything about the distribution, because these are not kind of natural things.

[00:45:32] Speaker 05:
They're humans making choices for all kinds of reasons.

[00:45:39] Speaker 05:
And here's a way that gets at the thing.

[00:45:41] Speaker 05:
But not simply by saying, Dr. Cohen did this, we claim his authority, and we're going to do it too.

[00:45:49] Speaker 00:
I'm not sure.

[00:45:49] Speaker 00:
Commerce, obviously, is looking to Dr. Cohen.

[00:45:52] Speaker 00:
And they're, I think, attempting to come up with a way of doing this that isn't necessarily the agency just from first principles attempting to reason something.

[00:46:03] Speaker 00:
But they are adapting and adopting something.

[00:46:05] Speaker 00:
They're not using Dr. Cohen's coefficient.

[00:46:07] Speaker 02:
Is that what they did here?

[00:46:09] Speaker 02:
Are you suggesting that's what they did, that's what happened here?

[00:46:12] Speaker 00:
Yes, I'm saying the way they have used, they're obviously not designing experiments.

[00:46:16] Speaker 00:
They're not designing social experiments, which is the context in which Dr. Cohen developed his coefficient.

[00:46:22] Speaker 00:
But they're trying to come up with, they're trying to adapt from him, what are rules of thumb for determining this first step of our analysis, just as they've established rules of thumb, like 33 and 66%, or a 25% difference for the other portions of the test.

[00:46:38] Speaker 00:
And those have been upheld as reasonable exercises of commerce's discretion.

[00:46:43] Speaker 00:
I see I'm out of time.

[00:46:44] Speaker 00:
So unless the panel has further questions, I'll retire.

[00:46:47] Speaker 04:
What about the currency exchange?

[00:46:50] Speaker 00:
Oh, the currency exchange.

[00:46:51] Speaker 00:
I thought we had forgotten all about that in our currency.

[00:46:57] Speaker 00:
No, we didn't.

[00:46:57] Speaker 00:
OK.

[00:46:58] Speaker 00:
This side of the bench is going to have to be.

[00:47:01] Speaker 04:
OK.

[00:47:02] Speaker 00:
All right.

[00:47:02] Speaker 00:
Well, yes, I know.

[00:47:04] Speaker 04:
So the other side pointed out

[00:47:08] Speaker 04:
in their briefing, look, we know what the rate exchange rate was.

[00:47:13] Speaker 04:
If you multiply it by a certain number, it gets you to another number, and that other number kind of neatly fits what really should have been the number.

[00:47:21] Speaker 00:
That's what they say.

[00:47:23] Speaker 00:
I don't know that we know what the right exchange rate was, because as commerce observed, you use this exchange rate in your calculation, but we don't know where it came from.

[00:47:31] Speaker 00:
You haven't documented the origin of this exchange rate.

[00:47:34] Speaker 00:
It, in fact, differs from exchange rates used elsewhere.

[00:47:38] Speaker 00:
It also appears to be not entirely relevant to the time period.

[00:47:43] Speaker 00:
And they go so far, as Judge Toronto pointed out, to say, well, okay, maybe we didn't really document this, but you should have taken independent steps to verify its accuracy using your own information that was at your disposal.

[00:47:57] Speaker 00:
That's an argument that's both waived and unexhausted, and for that matter, Marmon's auditors don't appear to have used this rate in their own conversions, which you can

[00:48:04] Speaker 00:
look at Appendix 3602 and Appendix 4679.

[00:48:08] Speaker 00:
That's not me attempting to back anything out.

[00:48:10] Speaker 00:
That's Marmon itself saying what exchange rate its auditor used in converting certain figures from its financial statements.

[00:48:18] Speaker 00:
And beyond that, when they're saying we need to make this adjustment,

[00:48:25] Speaker 00:
To me, it's very telling that they want Commerce to make an adjustment that would entirely offset their auditors' own adjustments to their audited financial statements as part of the reconciliation.

[00:48:35] Speaker 05:
Let me just see if I understand it.

[00:48:38] Speaker 05:
This is my understanding of their point.

[00:48:40] Speaker 05:
There's nothing strange about that at all, because the mistake was not the auditors.

[00:48:44] Speaker 05:
The mistake was their lawyers or whoever filed that first

[00:48:48] Speaker 05:
first filing in taking from the wrong place a US dollar figure for these things.

[00:48:58] Speaker 05:
So all they're trying to do is to correct their own mistake, the they here being not the auditor.

[00:49:04] Speaker 00:
Yes, that is their argument that this mistake is their own and has nothing to do with the auditors.

[00:49:10] Speaker 00:
But that is not, they at most have come up with their own potentially plausible reading of the record.

[00:49:18] Speaker 00:
that does not mean it supplants Commerce's reasonable view of the record under the substantial evidence standard, and that's what they would have to achieve here in order to prevail.

[00:49:28] Speaker 00:
They haven't established that, as Commerce has explained in its remand-redetermination and is further supported by the fact that the exchange rate they used in this calculation seems to not be documented and is contradicted by other evidence in the record as to what would have been potentially exchange rates applicable to the correct period.

[00:49:50] Speaker 02:
Thank you.

[00:49:56] Speaker 02:
We'll restore five minutes if you need.

[00:49:59] Speaker 06:
I'm sorry, your honor, how much?

[00:50:00] Speaker 06:
Five minutes or three minutes?

[00:50:02] Speaker 06:
Okay, thank you very much.

[00:50:05] Speaker 06:
First I'd like to say... I'm sorry, five.

[00:50:09] Speaker 06:
Oh, thank you.

[00:50:11] Speaker 06:
First...

[00:50:19] Speaker 06:
There was one issue I did not get to out of the interest in time.

[00:50:22] Speaker 06:
That has to do with averaging Marmon's reported product-specific play costs.

[00:50:28] Speaker 06:
I will just say that we're resting on the briefs.

[00:50:31] Speaker 06:
That's all I'd like to add.

[00:50:33] Speaker 06:
It's fine.

[00:50:35] Speaker 06:
Regarding the minor correction to the cost reconciliation,

[00:50:40] Speaker 06:
Just in brief, Commerce's reading of the record is not reasonable, and that's why it can't be sustained as supported by substantial evidence.

[00:50:48] Speaker 06:
None of Commerce's conclusions are reasonable, and in particular with respect to the exchange rate

[00:50:56] Speaker 06:
It is egregious, outrageous to reject a correction on the grounds that Momin failed to support an exchange rate when exchange rates are objective things that can be measured.

[00:51:08] Speaker 06:
And in fact, Commerce measures and records exchange rates on its website daily for purposes of anti-dumping margin calculations.

[00:51:19] Speaker 04:
Turning to Collins D. Just real quick, is it true that your reconciliation worksheet had entries

[00:51:25] Speaker 04:
in the worksheet that included auditor adjustments and yet for some strange reason for item L, in your view it failed to do that?

[00:51:37] Speaker 06:
The auditor's adjustment refers to the

[00:51:42] Speaker 06:
Maman's former accounting practice in 2018 when it did not convert U.S.

[00:51:48] Speaker 06:
dollar purchases to Canadian dollars in its system.

[00:51:53] Speaker 06:
So what happened?

[00:51:54] Speaker 04:
I'm just trying to find out the worksheet, the reconciliation worksheet itself that you turned in.

[00:51:58] Speaker 04:
Is it not true that there are entries in that worksheet that reflect the auditor adjustments?

[00:52:05] Speaker 04:
But in your view, for some reason, whatever reason,

[00:52:12] Speaker 04:
not also apply the auditor related adjustments there.

[00:52:18] Speaker 06:
Yes I apologize your honor I was just trying to build up to your question answer the answer to your question I'll get there quicker.

[00:52:26] Speaker 06:
The items you're referring to are items P, Q, and R.

[00:52:29] Speaker 06:
in the cost reconciliation, Excel lines 41, 42, and 43.

[00:52:33] Speaker 06:
Those do relate to the auditor's year end adjustment at the end of the year to make sure that all values in Marmon's audited financial statements are expressed entirely in Canadian dollars.

[00:52:44] Speaker 06:
However, as we stated clearly on the record, this correction to one line of cost reconciliation worksheet had nothing to do with the auditor's adjustment at year end.

[00:52:55] Speaker 06:
It was simply an inadvertent error by Mommy, taking the purchase value from 2018 and forgetting that, oh, we have to convert this purchase value to U.S.

[00:53:05] Speaker 06:
dollars.

[00:53:05] Speaker 06:
That's all it is.

[00:53:06] Speaker 06:
It's totally separate from the auditors.

[00:53:08] Speaker 05:
Was it the auditors adjustment on PQ&R for the purpose of the AD investigation submission or for internal business purposes?

[00:53:18] Speaker 06:
It's just the normal independent auditing process, having nothing to do with- As part of his audit?

[00:53:23] Speaker 06:
Yes.

[00:53:23] Speaker 04:
Okay.

[00:53:24] Speaker 04:
But as part of that audit,

[00:53:30] Speaker 04:
But for whatever reason, your side did not take that from the auditor adjustment and apply that to item L, even though it applied other auditor adjustments from that same year to other items.

[00:53:47] Speaker 04:
Is that a fair understanding of what happened here?

[00:53:50] Speaker 06:
not entirely, Your Honor.

[00:53:52] Speaker 06:
The auditor's adjustment is one global exchange rate calculation or conversion made at the end of the year for purposes of the audited financial statements.

[00:54:01] Speaker 06:
All that happened here, and it's really understandable, is that in preparing this cost reconciliation worksheet, when Momin

[00:54:09] Speaker 06:
It has to deduct Marmon Inc's purchases of wind tower sections from Marmon Energy.

[00:54:14] Speaker 06:
And in taking that value of the deduction, Marmon went straight to its accounting system, which in 2018 left that purchase value unconverted, left it in U.S.

[00:54:26] Speaker 06:
dollars.

[00:54:27] Speaker 06:
That's all that happened.

[00:54:28] Speaker 06:
At the year end, yes, the auditor would take all purchases, all transactions conducted in foreign currencies

[00:54:35] Speaker 06:
and convert them to Canadian dollars, including Marmon Inc's purchases of wind tower sections from Marmon Energy.

[00:54:42] Speaker 06:
However, this was Marmon's inadvertent independent error, just preparing a cost reconciliation worksheet.

[00:54:48] Speaker 06:
It went to its accounting system and not to the auditor's calculations for that one line.

[00:54:55] Speaker 06:
Cohen's D, what can I say?

[00:54:57] Speaker 06:
I don't have much time.

[00:54:57] Speaker 06:
But I will say that commerce uses it wrong.

[00:55:01] Speaker 06:
The Cohen's D coefficient does not have some.

[00:55:04] Speaker 06:
And the 0.8 threshold for large does not have some independent meaning.

[00:55:08] Speaker 06:
It only has meaning in relation to percentages of non-overlap.

[00:55:12] Speaker 06:
And those percentages that Professor Cohen defined

[00:55:19] Speaker 06:
will not be reached the same percentage as if the three assumptions are not satisfied.

[00:55:24] Speaker 06:
And the Michi Curia did a very good job of explaining this clearly at pages 17 to 19 of their brief.

[00:55:30] Speaker 06:
So I'd recommend ask the court to please review that.

[00:55:33] Speaker 06:
Thank you very much.

[00:55:35] Speaker 02:
Thank you.

[00:55:36] Speaker 02:
We thank both sides the case is submitted.