Reliably and repeatedly grading the color of diamonds

[Garry H (Cut Nut)]

you just lost 3 brownie points and a gold star Marty

We had a different agenda - but needed to make a decison about which stones to focus on for a grading survey.

Your distribution might look more like this after inclusion of ungraded stones and stones that are graded for retail sales rather than B2B trading.

But then you need to ask, in Bob Dylan style, who are you trying to be the ferryman for?
The 1st tier labs, or all labs?

I think you do not want to study all stones - I think you want the data Leonid provided - in fact - maybe only the GIA, AGS and maybe HRD listings?

 

[oldminer]

Maybe we can fly in an expert with a Gran to be an unbiased witness. Then we can run the diamonds on my SAS2000, their Gran and the ImaGem all in one day and in one place. If GIA graded diamonds are utilized, we don''t have to have anyone use human grading opinions as that is already a "given" with GIA reports.

The subject of color grading with all the possible varieties of color and UV fluorescence is very broad. We ought to stick with the yellow and yellow brown series of common colors probably.

 

[Richard Sherwood]

It''s really very simple guys.

We just want to know how often your machines agree with GIA or AGS grading.

If they disagree, how often and by how much?

A simple survey. Don''t get the humans involved, just the machines, and use the two premier labs in the world as the benchmark.

To keep it even simpler, I would agree with Dave in restricting it to just the cape series, and would suggest keeping it to medium fluorescence or less. I think it would also keep it simpler to restrict it to round stones, as fancy shapes introduce even more variables.

We''re not looking to demonstrate the wide world of color grading variables here. We''re just looking for a reasonable comparison which clearly demonstrates the grading of these machines in relation to major lab grading.

If the grading is done at Dave''s location, I would suggest an independent grader be there for operation of the Gran and SAS. The best would be a Gran representative and Marty.

 

[adamasgem]

Date: 11/21/2005 6:18:41 AM
Author: Garry H (Cut Nut)
you just lost 3 brownie points and a gold star Marty

We had a different agenda - but needed to make a decison about which stones to focus on for a grading survey.

Your distribution might look more like this after inclusion of ungraded stones and stones that are graded for retail sales rather than B2B trading.

But then you need to ask, in Bob Dylan style, who are you trying to be the ferryman for?
The 1st tier labs, or all labs?

I think you do not want to study all stones - I think you want the data Leonid provided - in fact - maybe only the GIA, AGS and maybe HRD listings?

Garry.. I don''t know what you did on the graph you showed, other than doctor MY graph in the lower color grades (i-J-K).

I don''t think that what you showed it is anything that Leonid did, as you implied, and you didn''t even renormalize the data.. My data was for the 1.0 to 1.49 ct ranges, and since your published data covered different size ranges other than what I looked at two years ago, i would suspect strongly that the "true" distributional statistics of available stones will (would be) different (but I might be wrong).

I suspect that there may be more relatively better colors, percentage wise in smaller size stones..


As to the use of HRD stones, I''m not sure the European grade ranges are the "same" as the US

 

[Garry H (Cut Nut)]

Marty - I was trying to help others understand what you are writting about. It was not an attempt to make definitve graphs etc - so calm down.

I really do not care if you do this survey at all - as i see it it has little to do with Pricescopes core purpose (cosmer advocacy) as it will have very little impact on the real world market.

The result will be good PR for 1 or 2 of the 3 companies, and bad for 1 or 2 of them.
It is, as Marty says Rich - probably not simple to do, and therefore the results could be meaningless - or might only appply to a very narrow range of variables that you''all decide to narrow down too.

So good luck. happy to help out - but do not appreciate being shot at.
You call me, as I will not be calling in here anymore.

 

[Richard Sherwood]

Date: 11/21/2005 6:29:41 PM
Author: Garry H (Cut Nut)
I really do not care if you do this survey at all - as i see it it has little to do with Pricescopes core purpose (cosmer advocacy) as it will have very little impact on the real world market.

The way I see it diamond grading is heading more and more in the direction of machine grading. Humans will never be eliminated, but machines will definitely begin to take a greater part in the process. It is inevitable.

I think consumers and dealers alike would be interested in which machines grade consistently close to major lab grading. I know my clients are always extremely interested in how the SAS's grade corresponds to the cert grade.

PriceScope did a first in comparing the three major labs. I feel that it would be another first for them to compare the three major color grading machines.

The result will be good PR for 1 or 2 of the 3 companies, and bad for 1 or 2 of them.

Was this not the same scenario you had going into the GIA-AGS-EGL analysis? And were the results not interesting in how fairly consistent the labs were, including EGL? Perhaps the same might be found in respect to the machines.

We've already got bad PR in the case of Dave stating the SAS-2000 grading was inferior. The comparison will give a chance for the facts to be known. Perhaps the SAS is inferior to the ImaGem. Perhaps the Gran will end up being a "sleeper" and deliver more consistent results than the SAS or ImaGem.

As a gemologist appraiser, I would find it extremely interesting. My guess is consumers would find it interesting as well, and you can bet that diamond dealers would be watching the results. If one machine consistently grades like the GIA, I would not be surprised to see a flurry of dealers purchasing that machine, or giving that technology their business.

It is, as Marty says Rich - probably not simple to do, and therefore the results could be meaningless - or might only appply to a very narrow range of variables that you'all decide to narrow down too.

Every day I decide on the color grade of diamonds. I analyze them, assign a grade and put it on a piece of paper. The GIA analyzes their stones, assigns a grade and puts it on a piece of paper. The machines analyze, assign a grade and prints it on a piece of paper.

In the end, everybody makes their decision and puts it in writing, human and machine alike.

What could be simpler? You just compare the pieces of paper.

or might only appply to a very narrow range of variables that you'all decide to narrow down too.

What "very narrow range"? If we decide on the cape series, we are encompassing 95+% of all diamonds sold. Keeping it to round brilliant cuts only further simplifies the results.

The problem we have here is personality conflicts, not procedural conflicts. This is an easy comparison which can be done at little cost and effort. If everybody can just swallow their egos for a while, we might get some productive info out of this which would clear up a lot of smoke and mirrors.

Marty...Dave...here's a chance to prove your technologies. It might possibly be the best thing that ever happened to both of you (or who knows, maybe Gran will kick your ass).

We know from the GIA-AGS-EGL comparison that this website can capture national and even international attention. If you can win Garry over and talk Leonid into doing what he does best, you might have an opportunity here.

 

[adamasgem]

Rich.. I''m behind you 100% on some sort of relative testing, and I know that you have been trying to build your own data base, and you should, and publish your results.If you notice, my software allows you to automatically build this sort of data base. Just click the buttion YES if you want to save an already graded stone. Dave appears honestly interested is presenting his results on the supposed thousands of GIA graded stones they are running one Imagem, and I think they will be surprised at the distributions they will likely find relative to a given machine''s calibration and the overlap or supposeded "error" rate they find, because in truth, they may be right or GIA may be right. They are certainly not going to "calibrate" each and every machine with thousands of stones, but I think their exercise is probably aimed at selecting a "set" of "masters". I certainly don''t have the funds to do such a thing, but rely on on my "set" of masters, GIA set #4981 to calibrate each system, and when I deliver a SAS2000 I run the the clients own "GIA masters". Each time I grade a stone I assign an uncertainty to the grade, and automatically set a minimum of 20% of a color grade range (it is probably much better than that) for an uncertainty and then add (sum the variances) to come up with the predicted color grade.

That said, let''s be simplistic for our general non technical audiance..

Suppose we have three labs (or machines) grading the length of a piece of wood, and they employ PERFECT graders, and their sole purpose is to say whether a length of wood is less, than or greater than or equal to 36 inches long.

Now each lab (or machine) is given a "yardstick" , blessed by G-D, in this case GIA, BUT each "yardstick" (H master) is of a different length, let say lab one''s yardstick is 33 inches long, lab two is given one that is 36 inches long and lab three''s yardstick is 39 inches long, that is +/- 10%.

Now is each lab is giiven the same 18" piece of wood (a G), they ALL are going to say that the piece of wood is shorther than 36".(G)

If they all test a pieice of wood 36.5 inches long, two of the labs will come up with the same answer (H), and one will ay it is shorter (a G), when in true fact it is a H)

If they are all testing a piece of wood 40 inches long they are all going to come up with the same answer (H).

BUT they are all "correct", because G-D (a/k/a GIA) said so, when he blessed thier three different master lengths of wood as an H Master

Now suppose those three "Labs" were all inside one "lab", sort of like a lab that has more than one set of "reference" lengths of wood. Both East and West Coast GIA probably have more than one set of "masters", given their volume, and each of their "masters" is, in reality different, and they internally would give different answers to to the same piece of wood.

Now when I was visually grading, and before t=I got the "AGS est colorimeter reading, I didt know that my "H master" was and AGS 2.15, 30% into the H color grade range, in other words , "my yardstick" was 46.8 inches long. so even if I was a "perfect grader", I would have has a 30% "error" rate. Now as I said before, I think the general visual "yardsticks" out there are statically probably between 42 and 30 inches long, and they vary from grade to grade.

Now it was suggested before that we do 40(31) stones in each color grade, if you were doing a machine to machine comparison. That is 320 stones for the D to K range, the colors that are typically graded. The logistics and shipping costs and insurance are prohobitive for each ''machine" to grade the SAME set of stones, let alone the TIME required to do the study.

If each "machine" were grading a different set of stones, each "machine" would have to grade MORE than 320 stones to get the same level of relative accuracy that one would get if they were grading the same set of stones.

Now I argued that 40 stones per grade was probaly too low, but that number will change depending on whether we grade the same "set" or not.

Now I may potentialy get down to the Philiadelphia area on paid business, I''ll be happy to bring my 17 stone master set down, because I know quite a bit about my set, (including some tests which I can''t unfortunately share becase of NDA''s ),and see what Dave''s Imagem says on my set of "masters", IF it will be allowed to be published, and I''d be happy to check Dave''s SAS2000 while I''m there. And if Imagem would pay my normal rate and expenses, I''d be happy to let them run my masters on the Imagem they are using for their multi thousand stone GIA test. I''m sure they would find the information useful.

I''m just a small fish in the pond, and as Dave says, don''t have the financial resources to do what I''d like to do for the industry. Contributions and research grants are always accepted, and if anyone wants me to do some specific funnded research and would be allowed me to publish it, I''d be happy to accept the assignment at greatly reduced rates

I''d like to say this publically on this thread, that Dave and AGS funded my expenses

, and I put in the time for free, to go to Denver, to do the study (which appeared in JCK) that showed definate radiation damage in Diamonds due to Postal Service Irradiation of the mails post 9/11 and the anthrax scare. (By the way, GIA kept insisting that it didn''t effect diamonds, and I got into a litlle arguement with them on that in Tuscon

, but they wouldn''t back down, even though they were wrong!) Dave has always supported my efforts where he could, regardless of the banter that sometimes appears. I still consider him a friend regardless, because friends can disagree and get miffed at each other, just like husbands and wives.

 

[Regular Guy]

Belated apologies again for interrupting and barging in here. Also, sorry for not pulling out my school books, but yes, Marty, your mention of my old friend "student t" gives me the easy cause to understand you know at least what I mean that the sample size of what probably is 31, in at least some circumstances, will suffice. Sorry that may not work here. Then again, and with appreciation for providing the theoretical underpinnings for your worries, you pose a modified and new challenge...relating to judging not a set of diamonds for sale, but a set of masters, themselves. Very possibly, your volunteering to make them available will be a unique offer, not repeated by others, and could make for a nice, though small test...proving what I''m not sure, but giving a relative idea of something.

 

[oldminer]

Thank's Marty. We're grown up enough to get along and do what's right. That's primary to going forward, cooperation.

I do wish to set the record straight on the multi-thousand diamonds being run on one ImaGem that is now taking place. This is not a calibration excercise. The machine was calibrated upon delivery to the site. It automatically self calibrates with some frequency, but does not require user re-calibration going forward. This is not a calibration exercise, but a test of how well the device works in a real lab environment. If (When) it shows it can do the job properly, the people testing it will adopt it for permanent use in lab facilities.

During the testing phase, the GIA grades, the user's visual grades, and the GIA grades all become data for ImaGem to utilize for further improvements to potentially be made. All technology advances. It does not sit still. Does anyone remember the first portable cell phones with the large battery box? I believe that once this test is completed, we will have a better product that is ready for further exposure and comparison. ImaGem is of the opinion that the color grading is working very well right now. I'll leave it in the hands of the testing folks...... We'll see what they think in the coming weeks. I want to encourage opinions based on factual evidence and will try to do the same with my own, as well.

 

[Richard Sherwood]

Okay guys, I''ll buy the beer.

Marty, thanks for that excellent explanation.

 

[belle]

Date: 11/22/2005 10:37:15 AM
Author: Richard Sherwood
Okay guys, I'll buy the beer.

reliably and repeatedly?
we might need an unbiased and independent assessment on that!

cheers

 

[Richard Sherwood]

Date: 11/22/2005 10:48:00 AM
Author: belle

Date: 11/22/2005 10:37:15 AM

Author: Richard Sherwood

Okay guys, I''ll buy the beer.

reliably and repeatedly?

we might need an unbiased and independent assessment on that!

cheers

Of course we''ll probably get in a debate as to the potency, purity and color of the beer.

 

[Rank Amateur]

I propose we do away with the traditional D to Z color grading all together and opt instead for a range of "Guinness" to "Coors Light".

 

[JohnQuixote]

What color is that?

Shiner Bock.

GIA would call it a Bass.

 

[belle]

Date: 11/22/2005 11:47:43 AM
Author: Richard Sherwood

Date: 11/22/2005 10:48:00 AM
Author: belle

Date: 11/22/2005 10:37:15 AM

Author: Richard Sherwood

Okay guys, I''ll buy the beer.

reliably and repeatedly?

we might need an unbiased and independent assessment on that!

cheers

Of course we''ll probably get in a debate as to the potency, purity and color of the beer.

there''s always room for debate. as long as their is an adequate supply sample available.

 

[aljdewey]

Date: 11/22/2005 11:47:43 AM
Author: Richard Sherwood

Date: 11/22/2005 10:48:00 AM
Author: belle

Date: 11/22/2005 10:37:15 AM

Author: Richard Sherwood

Okay guys, I''ll buy the beer.

reliably and repeatedly?

we might need an unbiased and independent assessment on that!

cheers

Of course we''ll probably get in a debate as to the potency, purity and color of the beer.

Naaaaaaaaaaaaaaaaaaah - let''s just drink it!

 

[Richard Sherwood]

You guys are a bunch of lushes...

Date: 11/22/2005 2:37:04 AM
Author: adamasgem
Rich.. I'm behind you 100% on some sort of relative testing, and I know that you have been trying to build your own data base, and you should, and publish your results.

Actually Marty I've been building data bases on colored stones, never thinking a data base showing the accuracy of the SAS's diamond grading would be needed, as I always thought of that as an established fact.

I'll start compiling an SAS to Major Lab color grade database though, so I'll at least have a track record I can personally produce for the SAS's color grading ability.

Here's a comparison of certed stones I've graded recently with the SAS:

1.23 Princess Cut.......SAS "F" (1.298 on 1.000 to 1.499 F range).......GIA "F"

1.53 Round Brilliant.......SAS "H" (2.265 on 2.000 to 2.499 H range).......GIA & AGS "H"

1.52 Round Brilliant.......SAS "I" (2.601 on 2.500 to 2.999 I range).......GIA "H"

2.00 Marquise.......SAS "D" (0.400 on 0.000 to 0.499 D range).......GIA "D"

1.14 Jubilee.......SAS "J" (3.321 on 3.000 to 3.499 J range).......GCAL "J"

1.57 Round Brilliant.......SAS "F" (1.239 on 1.000 to 1.499 F range).......AGS "F"

1.08 Round Brilliant.......SAS "J" (3.324 on 3.000 to 3.499 J range).......GIA "J"

The 1.52 which the SAS graded an "I" and the GIA graded an "H" had a strange medium green fluorescence, which the SAS filters out.

That kind of grading works for me...

 

[strmrdr]

what did you rate this one at using your master set?
1.52 Round Brilliant.......SAS "I" (2.601 on 2.500 to 2.999 I range).......GIA "H" ?

Thats why I was interested in having multiple human grading done.
Which is right gia or sas on that one?
Is it a boardline stone?
Lots of questions unanswered on that one.

 

[Richard Sherwood]

Date: 11/22/2005 9:03:57 PM
Author: strmrdr
what did you rate this one at using your master set?

1.52 Round Brilliant.......SAS ''I'' (2.601 on 2.500 to 2.999 I range).......GIA ''H'' ?

Thats why I was interested in having multiple human grading done.

Which is right gia or sas on that one?

Is it a boardline stone?

Lots of questions unanswered on that one.

Yes, it was a borderline stone. Notice that 2.6 is a very high "I" on the AGS scale (2.50 to 2.99 "I" range). I also gave it an "I" visually though. What was interesting about that is the client had brought in two 1.5''s at the same time (the 1.53 and 1.52 on the list).

Both were GIA graded "H", but the 1.52 (with green fluorescence) noticeably had about a half a shade more color when positioned next to the 1.53 against a white background under diamond grading lights. Even the client (a PriceScope lurker) noticed it, and the SAS definitely nailed it ("2.26 H" versus "2.60 I").

I don''t know, I just keep going back to Dave''s statement about the SAS being "another example of color grading badly". How much more accurate could you get than that?

That statement really bugs me. I guess it shows.

Love ya Dave. Hate the statement.

 

[strmrdr]

thanks Richard.
interesting stuff.
Im looking forward to the results of the study.

 

[oldminer]

Rich; Your SAS is doing a better job than mine, I''m afraid. I''m glad it is doing better for you. Obviously the use of the word "badly" was poorly chosen. I might have said that the one I own does not grade well enough for the purposes I initially thought I would use it for. As I have said, we do use it for things where the functionality is excellent. If I got results like you were getting consistently, then I''d be happier with it than I am. I''d then be in your camp.

Better?

 

[Richard Sherwood]

Yes, much. Thank you Dave.

 

[adamasgem]

Leonid has provided me some "availability" data for 4 labs for the 1.0 to 1.49 ct range, and I'm tryng to write up my interpretation, but I want to wait for a larger sample size and cover all shapes and sizes. In the meantime, let me address the fluorescence issue http://www.adamasgem.org/giafluor.html , where historically "diamonds should be graded at their poorer color in artificial light devoid of ultraviolet. in a large sample set, if their were no effect on grading color because of fluoresecence, one would expect that the color distributions would be independent of fluorescence. Well I spent some time today on a database I have access to and downloaded and separated some very interesting information, just on GIA stones. We all have heard about the GRAN regarding fluorescence, and statement that the SAS2000 tries to filter the UV content out, but the data below suggests that fluorescnet stone might not be equally graded, and I can enumerate several reasons why in a court of law that would pass any scientific muster, as well as easily demonstrate. That is why I have been doing some thinking about sample size, type of stone etc, to sort of even out the playing field.

Now GIA does not specify the COLOR of fluorescence when they say NONE or Faint, like they do when the fluor is above blue, but the vast majority of stone have blue, and I have seen numerous examples of "NONE" where it approached medium blue, and that is why I, along time ago, because I (and other appraisers I know) got sick and tied of explaining that the meaning of "NONE" is sort of like the "depends on the meaning of the word IS" is" to quote a famous individual, i had filed an FTC complaint agianst GIA, with zero results.

I don't want to spend much time with this, so I won't be commenting unless valid questions come from those with technical credentials in math or science regarding the statistical significant of these data, but I think the data speaks for itself. There can be many "arguments" made that "availability" doesn't reflect reality, but "availibility" is where we would draw draw any sample set for a valid test of differences from some "standard", and I've long thought something smells with grading..

I'll enjoy reading reading your comments on this post..

PS I didn't include the VERY STRONG BLUES because there were only about 150 total in the sample set

 

[perry]

Marty:

I saw you were posting and thought I would drop in on the discussion. The graph is interesting, and I have to but remember all the statistics I was taught (but I still work with math and statistics at work - something few other engineers really do).

What is the verticle portion of the graph: It cannot be fraction of total - othewise the three lines would add up to 1.0. Nor do I think it is percent.

Let me know, and I''ll let you know if the twitch in my nose seems real.

Perry

 

[adamasgem]

Here is the same data in a bar chart with decimals multiplied by 100 to indicate percentages. You use the x-y format to try to estimate the type of funtional relationship for a continuous probability distribution function.

 

[valeria101]

Date: 11/23/2005 10:11:48 PM
Author: adamasgem

in a large sample set, if their were no effect on grading color because of fluorescence, one would expect that the color distributions would be independent of fluorescence.

True enough.

I would surely love to see these results and the data too, if my practice with econometrics qualifies for it. That is not exactly training in exact sciences, so... feel free to ignore this if irrelevant. Just ''thought they are better said than not with the risk of sounding silly.


Assuming I got it right that the tables show the frequency of fluorescent items out of each color group... one argument against the findings might come from having smaller samples for the lower colors. But this is an ''availability'' argument I am not fond of either. If only my usual econ data would be half this good...

Otherwise, could it be that:

#1. the distribution of fluorescence across color grades is not random in the rough ? If there is some correlation between the physical properties producing color and fluorescence, that may need to be accounted for

#2. does body color interfere with the observed degree of fluorescence in some objective way?

#3. if the presence of fluorescence causes random grading error, it would not be captured by the regression suggested in the chart, while still causing problems with the classification of the affected stones. GIA published survey on fluorescence found inconsistent identification of fluorescence - at least to some degree. Perhaps that panel study was not related to grading inconsistencies (sure that!) and it could have well been unfounded etc. Logically, fluorescent stones should be over graded, but why if they are not consistently over graded because graders do observe fluorescence and try to compensate with random success?


With data at hand, these

could be tested to some extent.

 

[adamasgem]

Date: 11/24/2005 1:00:19 AM
Author: valeria101

Date: 11/23/2005 10:11:48 PM
Author: adamasgem

in a large sample set, if their were no effect on grading color because of fluorescence, one would expect that the color distributions would be independent of fluorescence.

True enough.

I would surely love to see these results and the data too, if my practice with econometrics qualifies for it. That is not exactly training in exact sciences, so... feel free to ignore this if irrelevant. Just ''thought they are better said than not with the risk of sounding silly.


Assuming I got it right that the tables show the frequency of fluorescent items out of each color group... one argument against the findings might come from having smaller samples for the lower colors. But this is an ''availability'' argument I am not fond of either. If only my usual econ data would be half this good...
The data represents the distribution of stones graded by GIA only, our "reference", if you want to put it that way, Scientifically, the relative position of a fluorescent stone to the UV conatianing light source effect the percieved color. These data may or may not represent all diamonds, as most are junk, and represents the IF to SI2 clarity crades, as GIA seems to grade very few, percentage wise of the Included catatgory, I feel there is a statistically significant difference in the dustributions given the fairly large sample sizes..

Otherwise, could it be that:

#1. the distribution of fluorescence across color grades is not random in the rough ? If there is some correlation between the physical properties producing color and fluorescence, that may need to be accounted for

Which came first, the chicken or the egg, is you dont have UV, you will get a different body color in a fluorescent stone. GIA has changed horses in mid stream. Look, a fluorescent tube UV output is not connsistent from the center to the edges, and there is a 1/r to 1/r^2 falloff in the UV as you move away rom the tube, and each bublb produces differert amounts of UV, and GIA has changed bulbs types and manufacters over the years and each grader''s technique os different.. read my article

#2. does body color interfere with the observed degree of fluorescence in some objective way?

Self absorption is a factor as a function of wavelength..

#3. if the presence of fluorescence causes random grading error, it would not be captured by the regression suggested in the chart, while still causing problems with the classification of the affected stones. GIA published survey on fluorescence found inconsistent identification of fluorescence - at least to some degree. Perhaps that panel study was not related to grading inconsistencies (sure that!) and it could have well been unfounded etc. Logically, fluorescent stones should be over graded, but why if they are not consistently over graded because graders do observe fluorescence and try to compensate with random success?

technique, etc effect the grade, age of the bulb, position on the tray, lots of factors. FOR over 50 years they said no or minimal UV and then they show the graders holding the stones right up to the light with no mention of eleiminating or having minimal UV content. It is all about "better" paper and $
With data at hand, these

could be tested to some extent.

the data suggests an inconsistensy in grading, and a shift in color towards better colors in fluorescent stones, something we KNOW and can show happens.

I don''t think a test with GIA grades as the "truth" which uses fluorescent stones is necessarily valid, Period, and it distorts relative machine accuracy by throwing another degree of freedom into the mix. It really pays $ to have a E-IF rechecked to be a D-IF.

I''d love to see data like this from 30 or 40 years ago...I feel it is a problem..This is a fairly large sample set, probably larger than most "polls" or "surveys" people treat as "truth" politically.

I''d like to see IGI stats but the sample data available is next to nothing compared to GIA paper. We will show some ineteresting data later....

 

[oldminer]

Marty;

SInce we know many GIA stones graded as "none" have quite a bit of visible fluorescence, would that have an effect on the graphs you made from the Pricescope database? It seems to me that a D color "none" is more rare on the graphs than in reality. oO you agree?

How would the graphs look if you predicted the distribution knowing what you know about actual occurrence of UV fluorescence?

 

[Serg]

re:It seems to me that a D color "none" is more rare on the graphs than in reality. oO you agree?


In reality a few seller want grade in GIA J-L diamonds. In reality( all world) D diamonds are more rare, J-L are more world-wide.

 

[adamasgem]

Date: 11/24/2005 9:10:42 AM
Author: oldminer
Marty;

SInce we know many GIA stones graded as ''none'' have quite a bit of visible fluorescence,{gee, and I thought none was supposed to mean none

} would that have an effect on the graphs you made from the Pricescope database? It seems to me that a D color ''none'' is more rare on the graphs than in reality. oO you agree? Yup

1) I used a different database for these data, Leonid is working on automating the search on PS
2) I have''nt Published lab to lab differences, but one could infer that from the data I, and Leonid have done independently. The sample sizes on other labs are 15% or less than that of GIA in total, and the the AGS data certainly impies that, but I''m trying to get ahold of a bigger AGS sample set, so as not to unintentionally confuse the public. But what I found in total, consistent with Leonid''s search on 1ct rounds, is that surprisingly the AGS and EGLUSA data is markedly different than GIA''s when it comes to D''s, but I wouldn''t want to speculate regarding GIA''s distribution, given the current scandal. Now if you averaged the AGS and EGLUSA data, it would certainly suggest that GIA gives "better" grades.


How would the graphs look if you predicted the distribution knowing what you know about actual occurrence of UV fluorescence?

Well, in the G&G 1997 article on fluorescence, GIA stated that 35% of their data base "REPORTED" some fluorescence. The "Faint" range data and the "none" range distributional data for GIA actually overlapped, so I combined them, as I couldn''t isolate "faint range" blues from other colors. And in the data base I searched, there was a substantial difference between the total GIA stones and the total of those showing definate comments "none", faint, medium, strong and very strong, as well as there were some time "dynamics" in the totals of the database as I was searching it.