Future Certification guidlines {wink}

[Kalkbreath]

I heard that

MAC will be testing twelve fish a year ......or one fish a month as part of their new cyanide detection certification. ......... the backers of this new testing format are pleased that the industry reeform movement is behind such testing and the industry has ut most confidence in this type of "intense" screening of livestock's {not scene sinch the late 1990s!. The people responsible performing the tests have stated the desire to test all twelve fish for the year within the first calendar month {Jan} ........ One, so that the public can be certain a head of time that this years catch is certified and second, so that all tested fish can be collected by the same diver .

: wink

 

[Kalkbreath]

Has any one else heard this ? Sounds fishy to me ?

 

[mkirda]

I heard that

MAC will be testing twelve fish a year ......or one fish a month as part of their new cyanide detection certification. ......... the backers of this new testing format are pleased that the industry reeform movement is behind such testing and the industry has ut most confidence in this type of "intense" screening of livestock's {not scene sinch the late 1990s!. The people responsible performing the tests have stated the desire to test all twelve fish for the year within the first calendar month {Jan} ........ One, so that the public can be certain a head of time that this years catch is certified and second, so that all tested fish can be collected by the same diver .

: wink

Hey, kalk.

It is great to see you quoting yourself spouting such BS.

Makes you look even more foolish than normal.

As for the insanity of testing 12 fish per year, the only source of this information is your own posts! Don't you get dizzy being so self-referential?

For your reference, YOU ALONE are advocating sample sizes that would result in error bars in excess of 30%. Saying that you are 95% confident that mean percentage of cyanide caught fish is 25% plus or minus 30% is essentially meaningless.

ONLY YOU advocate such foolishness.

Regards.
Mike Kirda

 

[Kalkbreath]

You all agreed that 750 fish was all that one needs to gauge five years of collection from PI [15 million fish }..........Not me ! Funny how the tide turns when your confronted with your own notions :wink: Your busted

 

[mkirda]

You all agreed that 750 fish was all that one needs to gauge five years of collection from PI [15 million fish }..........Not me ! Funny how the tide turns when your confronted with your own notions :wink: Your busted

Show me where I am quoted stating that.
I never said such a thing.

 

[mkirda]

You all agreed that 750 fish was all that one needs to gauge five years of collection from PI [15 million fish }

Ok, Kalk... Which is it?

First you quote 13,000,000 fish per year.
Then you quote 14,000,000 fish per year, which is what I used to base the top number of my chart on.

Now you are stating 15,000,000 over five years. i.e. 3 million per year.

Which is it? I can't keep your made-up numbers straight when you keep jumping around like this.

Anyway, I'm getting good at this calculation now...

95% confidence level, 5% error bars, population of 3,000,000:
Sample size required: 384 per year.

Interestingly, the same sample size is required for a population of 5, 13, or 14 million.

If you disagree, show me your math to come up with a different answer.

 

[Kalkbreath]

Well............. then I have good news and bad news ......Peter did not come close to your new minimums....750 fish for five years at about five million fish a year. {PI sold more fish back then because there were few other Countres to buy fish from } Thats 750 fish sampled from 25million . The good news is that to certify fish under your minimums .....MAC only has to test about 100 fish a year for every 50,000 they collect. And they get to eliminate eighty percent of the samples.....{like Peter and the missing 4000 fish from the 4800 fish in the original data] :lol:

 

[mkirda]

Well............. then I have good news and bad news ......Peter did not come close to your new minimums....750 fish for five years at about five million fish a year. {PI sold more fish back then because there were few other Countres to buy fish from } Thats 750 fish sampled from 25million .

Well, I have some good and some bad news too.

First the good news. 384 fish is all that is needed when N=25 million as well,
again with 95% confidence and 5% error bars. So 750 would be sufficient.

Bad news: 48,000 fish sampled is more than sufficient. A sample size of 4160 is all that is needed when N=25,000,000, with a confidence rate of 99% and error bars of a mere 2%.

In other words, your statements are incorrect.

The good news is that to certify fish under your minimums .....MAC only has to test about 100 fish a year for every 50,000 they collect.

You have yet another math error, Kalk. You really should look up that equation...

N=50,000 needs a sample size of 381 to reach a 95% confidence level with 5% error bars.

For the record, for N=50,000, a sample size of 96 would give you a 95% confidence level, but the error bars would be 10%. That is not good enough, Kalk.

 

[Kalkbreath]

Well............. then I have good news and bad news ......Peter did not come close to your new minimums....750 fish for five years at about five million fish a year. {PI sold more fish back then because there were few other Countres to buy fish from } Thats 750 fish sampled from 25million .

Well, I have some good and some bad news too.

First the good news. 384 fish is all that is needed when N=25 million as well,
again with 95% confidence and 5% error bars. So 750 would be sufficient.

Bad news: 48,000 fish sampled is more than sufficient. A sample size of 4160 is all that is needed when N=25,000,000, with a confidence rate of 99% and error bars of a mere 2%.

In other words, your statements are incorrect.

The good news is that to certify fish under your minimums .....MAC only has to test about 100 fish a year for every 50,000 they collect.

You have yet another math error, Kalk. You really should look up that equation...

N=50,000 needs a sample size of 381 to reach a 95% confidence level with 5% error bars.

For the record, for N=50,000, a sample size of 96 would give you a 95% confidence level, but the error bars would be 10%. That is not good enough, Kalk.

Peter only included 7500 of the 48000 fish......will MAC also be alowed to omit seventy percent of the data? Will MAC be alowed to average out the data? without revealing how much of the data was included for each year? A study with most of the 7500 fish collected in 1996 would weight the findings now would it not ? And will MAC be alowed to only include five to ten fish from each species per year? Like eighty percent of Peters data?.........Lastly when MAC tests their 96 fish each year.........I dont want to hear you and others crying that the sampled MAC fish dont reflect the ratos of the collected fish ..... :wink:

 

[PeterIMA]

Kalk, You implied that most of the samples on aquarium fish were taken in 1996. Actually, there were 601 fish tested in 1996, 570 in 1997, 3167 in 1998, 2128 in 1999, and 2681 in 2000 reported in the published CDT paper.

You have assumed that within each Family that the species were sampled equally. Hence, in the case of the Pomacentridae you assumed that there were only about 4-5 fish of each species (111) sampled per year. This was not the case.

It is a fact that rare species were not sampled very much and that common species were sampled at a higher rate (to some degree in proportion to their abundance). This actually served a purpose, it allows for robust numbers of fish sampled on the species that were most frequently exported. I was quite interested in the findings by Blue Hula that the proportion of fishes found to have cyanide present (25%) did not change significantly, if you exclude the rarer families (and rarer species in those families) from the calculation (remains 25%). Hence, the sampling was representative even when you excluded the rarer species groupings.

It is of interest to ask how will the MAC do it, if they somehow take over the responsibilities of BFAR? Will they conduct random sampling across the country (PI)? How should this sampling be conducted? How will we know they are doing this in a scientifically rigorous fashion? Will it be tied to law enforcement (or will they just post results on a web-site to give the trade a lesson with a wet noodle)? Assuming it is used for prosecution who should be prosecuted (A. Collectors, B. Middlemen, C. Exporters, D. Cyanide Dealers/Smugglers, E. All of the preceding)?

As far as sampling is concerned I have suggested that it could be randomized within Families (e.g. the sampler would take one fish species per family each time he/she sampled an export facility). The species sampled within each family would be randomly chosen. Will the MAC do this? Will BFAR do this? I think we have to consider that this will actually yield fewer of the most commonly exported species being tested for CDT. This makes for good science, but apparently is not what you want. If you think about it carefully, you might conclude that the way IMA did it (randomly by species without regard to Families) actually may be the best way to do it, from the point of view of regulating the aquarium trade. I solicit opinions from everyone lurking here. How should it be done?

Are there ways to "improve the process" of collection, transport, and CDT sampling and testing for a more sustainable trade? Please reply soon. Your comments can be on any of the topics (net training to fish mortality). I need the information for a panel discussion that will be held at MO 04 Conference. PS-Paul Holthus (MAC) will be on the panel along with Steve Robinson (AMDA), Ferdinand Cruz (EASI), Dante Dalabajan (EASI), Dr. Vaughan Pratt (IMA), myself (IMA), and Ambrosius Ruwidrijarto (Telapak).

Peter Rubec

 

[mkirda]

Peter only included 7500 of the 48000 fish......will MAC also be alowed to omit seventy percent of the data? Will MAC be alowed to average out the data? without revealing how much of the data was included for each year? A study with most of the 7500 fish collected in 1996 would weight the findings now would it not ? And will MAC be alowed to only include five to ten fish from each species per year? Like eighty percent of Peters data?.........Lastly when MAC tests their 96 fish each year.........I dont want to hear you and others crying that the sampled MAC fish dont reflect the ratos of the collected fish ..... :wink:

"Rationality is the recognition of the fact that nothing can alter the truth and nothing can take precedence over that act of perceiving it." :wink:

 

[hdtran]

Let me sort of quote myself with a hypothetical procedure:

(1) All fish being exported are bagged individually (presumably the way it's done now)
(2) All individual bags have a bar code (adds $0.01 per bag)
(3) A computer program chooses randomly 384 numbers out of the bar codes (adds $25K per year for computer, bar code scanners, divided by N fish) per year
(4) If the bag scans "yes" for testing (unknown to collector until it's at the warehouse), the bag is pulled out and tested per CDT procedure.
(5) If the fish tests negative, continue as usual, but
(6) If the fish tests positive, close down all exports.

As a consumer, I'd be willing to pay a higher price for fish, if the procedure is as outlined above. I'm willing to accept the risk that up to 5% of the collection would be unclean, and slip by the sampling procedure.

The beauty with random sampling (SPQC, etc) is that it does work. That's why your 2000 model year car is far more reliable than the 1973 model year. You don't need to determine "how many damsels, how many puffers." You never know whether you're about to test a damsel, a mandarin, or a tang. But in car-speak, you know that if a spot check reveals even a paint defect, the assembly line is shutdown, and then, everyone (temporarily) loses profit until the problem is fixed.

I'd be happy (on a PM basis) to run numerical simulations for anyone who's interested. Tell me how many hypothetical damsels you're exporting, how many hypothetical tangs, and how many hypothetical angels. Also tell me, of the hypothetical fish, how many are hypothetically unclean. I'll have the computer program (written in MATLAB, for those of you interested) generate the numbers, pull out 384 samples, and give you how many positives it pulls. I'll even give you my code for you to run yourself, and point you at some alternative freeware.

Regards,

Hy

 

[PeterIMA]

Hy, I think this could work if the fish were placed in "breathing bags" or some other container (like a plastic container with holes) so that the fish never came out of the container at each step of the chain. Then, tags could be used not only to control the CDT sampling, they would allow each bag/container to be tracked from reef to retailer. This would also allow one to determine mortality through the chain and determine which collectors had the best fish (that live) or worst fish (those that died the most). Tracking fish is supposed to be one of the MAC goals. Is it not?

 

[Anonymous]

So now all fish will get nose and fin burns from said cup, starve from no food or pop the paper thin breathable bags. Whats that going to solve Peter? Now we have to acclimate each fish in each bag or cup, hold all fish in bags or cups (gauranteed death BTW) and feed the fish in each cup or bag? I want you to come out here Peter, and try this for yourself. We're small, so it shouldn't be that hard, right? Science is so FAR removed from reality in this case, its not even funny.

 

[dizzy]

Peter,
If the plastic container had holes all the water would probably get sloshed out during shipping.

Gresh,
I wasn't aware that the fish got fed much until they landed at retail. I have heard some wholesalers feed the ones that get held over the weekend, but that was about it. Otherwise their waste would foul the shipping water. Do you guys feed then ship? How long do you hold before you ship out?
Mitch

 

[hdtran]

Are fish shipped multiple fish per bag, or one fish per bag? Or are they shipped multiple instances of one species in a bag (e.g. 10 blue damsels in bag 1, 10 yellow tang in bag 2, etc.)?

I'm sure a professional statistician could give you a good sampling scheme if it's multiple fish to a bag. I was giving you a proposition for one fish per bag. If you wish me to simulate numerically multiple fish per bag, I think I can write a program to do that pretty easily.

Hy

 

[dizzy]

Hy,
Most fish are shipped one to a bag. However damsels sometimes come two or more in the same bag. Sometimes they are both or all dead though.

 

[Anonymous]

hdtran - Amount of fish in a bag typically depends on what country of origin it is. Most imports are a single fish to a bag.

Gresh,
I wasn't aware that the fish got fed much until they landed at retail. I have heard some wholesalers feed the ones that get held over the weekend, but that was about it. Otherwise their waste would foul the shipping water. Do you guys feed then ship? How long do you hold before you ship out?
Mitch

I'll put it this way, we're on our second 5 gallon bucket of food for the last 3 months. I'm sure most major wholesalers don't feed much, if at all. We're smaller, so we can baby our fish easier then the big guys. Some fish couldn't even be reshipped if we didn't get them feeding again prior to shipping. Feeding is a selective thing, we typically know the next days ship outs, so we don't feed whats going out the next day. Amount of food is another consideration that point. In the wild, they consume major amounts of food, so they poop major poops (why they need to be cleaned out prior to shipping). We're not dumping pounds of food into the tanks, it's a controlled feeding. Your fish you recieved this week were all fed Saturday 3 times. You tell me, how'd they look? Were they're big black poops floating about? How long we hold a fish is dependant on the fish's health. Some fish land a lot better then others and are ready for sale within days. Science has to take a back seat on these types of considerations. There is no formula that can be followed that shows how long each fish need to be held, fed or packed. It can only be able to be done thru hands on learning (trial and error).

 

[Anonymous]

Dizzy, are talking about export shipping or within the states shipping? We NEVER have ANY damsels land gang bagged from any exporters.

 

[hdtran]

I'm talking I/P (Manila Int'l, Singapore Int'l, etc.) to LAX. Are the fish in that 747 cargo hold 1 fish/bag, 20 bags/box, or are we talking 2,3,4 fish/bag? Because what I'm proposing is not sampling at LAX, but sampling outgoing from Manila Int'l, randomly taken using a random number generator.

Here's a quickie MATLAB script that I wrote. It's going to be meaningless to most of you, since I'm a geek (and youse guys ain't 8) )

% n samples taken out of N fish...
% Simulation written in 'honor' of Kalkbreath
% This script will run in almost any version of MATLAB (www.mathworks.com)
% This script should run in almost any version of OCTAVE (www.octave.org)
% Script written by Hy D. Tran, PhD, PE; Feb 5, 2004 on MATLAB R12, classroom license
%
Nfish=1000000; %Number of fish, say 1000000 (one million); integer value
cnfrac=0.1; % Fraction with Cyanide, for 10 percent, enter 0.1; must be between 0 and 1
testnumber=384; %How many do we want to test?, say 384; integer value
%
%You can only modify the above numbers. Don't muck with anything after this line!
%
Popfish=rand(Nfish,1); %Create the vector Popfish with uniform random number bet. 0 and 1
Popcn=Popfish-(1-cnfrac);%We want a vector with a '1' if the fish is positive, '0' if negative
Popcn=gt(Popcn,0); %Turn a positive into 1, rest are zero
%Popcn is the population with cyanide, 1 if positive, 0 if not
SampleDummy=rand(Nfish,1); %Sampling vector
SampleDummy=SampleDummy-(1-testnumber/Nfish); %Won't take actual number of samples, but close to it
SampleDummy=gt(SampleDummy,0); %If SampleDummy()=1, take a sample, if not, don't...
NP=SampleDummy'*Popcn;%Take samples...
NSamples=sum(SampleDummy);%Number of actual samples taken=?
NSamples
NP
'% of samples positive is'
100*NP/NSamples
% End of script

If you're semi-geekish and cheap, you can download and install Octave on most Windows machines, and all Linux machines. Run this script, and it will tell you how many simulated fish were sampled positive. I've run the script several times, and get between 7% and 12% very reliably for 10% tainted fish. I don't need to do it, because I believe the math (I just don't remember the derivation of student-T). I doubt that it will convince Kalk, but I'd be happy to run this script with 10 million fish (takes about 5 seconds to run 1 million on my computer), sampling about 384 fish (I'm too lazy to code it to run exactly 384 samples, as you can obviously see).

Regards,

Hy