Fuzzy numbers?

[PeterIMA]

Horge, I think one can question the % presence for pufferfish based on the low sample size tested (30). Perhaps if more pufferfish were sampled and tested the result would be different. Please note my comments posted earlier today about the need for random sampling by species within families to equalize sampling between families.

As far as the lionfish the sample size is higher (93) so it looks more credible (even if Columbo can't figure it out).
Peter

 

[PeterIMA]

Horge, I think there is evidence in the CDT database that schooling species (such as tangs and caesionids) are being targeted by cyanide (despite the fact that they don't generally hide in the corals). This is an important fact that may be related to the dead food fish trade (that Kalk has referred to). In other words, fish are being caught with cyanide for: a) the live food fish trade (generally groupers, snappers, and Napoleon wrasse), b) the aquarium fish trade, and c) dead fish for consumption in the Philippines (dead fish trade. The latter could be a threat to human health, since fishes killed by overdoses of cyanide do not excrete the cyanide before the fish is consumed by humans (e.g., Filipino or Indonesian citizens).

 

[dizzy]

Peter,
I ain't horge. Horge has a much better grasp of the English language than I do. We do resemble though. :lol:

 

[mkirda]

Did you explain the math behind three fish being tested every five years and how these three fish wold reflect the average?

Yes.

Data is data is data. Go look at your Stats book, look up how to calculate Standard Deviation and Standard Error. Look how large your Standard Error is on the sample size you want to use as your example. The Standard Error tells you how reliable or to what degree of probability, that the mean is going to be accurate. You can add, subtract, multiply and divide, right? The toughest thing is the square root in the formula. You can figure it out for yourself just much confidence you can have in the mean with a data set size of your choosing.

You can do it. It ain't hard. You would then have the exact answer you seek.

Once you do go figure this, I then want you to cite where Peter ever said anything that overstated the mean. That will be the only 'proof' that will hold any water with me.

 

[mkirda]

What does the chart reflect on the IMA website ........You list thirty species of Blennies.......and 79 fish tested within those thirty species.......thats less then three fish for each of the 30 species. Which means less then one blenny per species was included in the study per year {five years}Most of the fish species in your data have less then ten individual fish included in the data. How is it possible that so few fish can reflect accurately the industry? It may very well be that three fish in your data will be the sole representation for a species that is imported by the tens of thousands,from over one hundred separate islands and a thousand individual collectors. How can three fish reflect all the different collection conditions? if all three fish in your data came from one collector, that one collector would reflect all the collectors that collect that species!

God, it is just not sinking in, is it?

Peter already explained this one to you, Kalk.

Proper statistical analysis requires random sampling and a data set.
The more elements in the data set, the more statisitically significant the mean will be. If you want the mean to be reliable to within say 3%, you need sample sizes to be rather large... Let's say 1800 or so. Look at your stats book, look for the equation (P<=0.05). That says that the mean is 95% accurate. Read up on it- You can see what sample sizes are required to get whatever percentage accuracy you require, given that you know the population size too. This isn't rocket science- it really is just basic math.

Because the sample sizes were so low for some species, Peter lumped the species into families, then ran the numbers for families rather than species. You lose precision on a per species basis, but due to sample sizes, you get a much greater increase in confidence in the mean for the family instead.

So there- We've explained the same thing to you twice now.

CAN WE MOVE ON?

 

[Kalkbreath]

Uncloaking briefly...

Kalk,

If you're looking at http://www.marine.org/Content/CDT/imgCDT/slide17.html and http://www.marine.org/Content/CDT/imgCDT/slide18.html,
then, your assertion that

Even by Peters data.....70 percent of the fish coming out of PI ARE netcaught! And More then likely 80% to 90% are clean fish

doesn't hold water.

As a....personal favor.....to me....would you....please....{post} without....ellipses?

Thanks!

Recloaking,

Then stay uncloaked long enough to explain why it doesn't? Keep in mind that ninety percent of the species in this data are rarely exported for the trade. About 70% of the fish imported { by number} to the USA are from only thirty so species. There are about one thousand species in the data. If we exclude all the oddball fish and only count the fish that are collected in great number........the clownfish , damsels, mandarins,chromis etc. We arrive at less then the 25% average.But thats hardly the point .........there are far too few fish in the data..........The truth could be that ninty percent of the exported fish contain cyanide!There is no way Peters data would reflect the truth?

 

[Kalkbreath]

What does the chart reflect on the IMA website ........You list thirty species of Blennies.......and 79 fish tested within those thirty species.......thats less then three fish for each of the 30 species. Which means less then one blenny per species was included in the study per year {five years}Most of the fish species in your data have less then ten individual fish included in the data. How is it possible that so few fish can reflect accurately the industry? It may very well be that three fish in your data will be the sole representation for a species that is imported by the tens of thousands,from over one hundred separate islands and a thousand individual collectors. How can three fish reflect all the different collection conditions? if all three fish in your data came from one collector, that one collector would reflect all the collectors that collect that species!

God, it is just not sinking in, is it?

Peter already explained this one to you, Kalk.

Proper statistical analysis requires random sampling and a data set.
The more elements in the data set, the more statisitically significant the mean will be. If you want the mean to be reliable to within say 3%, you need sample sizes to be rather large... Let's say 1800 or so. Look at your stats book, look for the equation (P<=0.05). That says that the mean is 95% accurate. Read up on it- You can see what sample sizes are required to get whatever percentage accuracy you require, given that you know the population size too. This isn't rocket science- it really is just basic math.

Because the sample sizes were so low for some species, Peter lumped the species into families, then ran the numbers for families rather than species. You lose precision on a per species basis, but due to sample sizes, you get a much greater increase in confidence in the mean for the family instead.


So there- We've explained the same thing to you twice now.

CAN WE MOVE ON?

Are you really stating that three blue damsels collected during a five year time frame will reflect the other 800,000. Other blue damsels? If so then lets test three more, if the test is off by even one fish coming up different then the original three fish tested by Peter........the result will be a thirty percent change in the data. One fish! One fish can change the results by thirty percent on that species!

 

[hdtran]

OK, I'll bite. Kalk, I'm uncloaking again, because my curiosity is killing me.

Kalk writes:

Then stay uncloaked long enough to explain why it doesn't? Keep in mind that ninety percent of the species in this data are rarely exported for the trade. About 70% of the fish imported { by number} to the USA are from only thirty so species. There are about one thousand species in the data. If we exclude all the oddball fish and only count the fish that are collected in great number........the clownfish , damsels, mandarins,chromis etc. We arrive at less then the 25% average

Which 30 (or so) species encompass 70% (or so) of the trade? I didn't spot you stating them earlier in this thread (but then, the ...'s were distracting me). Can you also give me a breakdown by percentage? E.g. how many %chromis, %clowns, %mandarins, etc., totalling to 70% of the imports? By the way, I assume that all the Chromis & Clowns are lumped together in the Damsels, as they are all classified in the same family. Or, if you don't want to type (retype) the breakdown, would you be so kind as to post a citation (preferably online, but paper is fine)?

Thanks!

 

[mkirda]

What does the chart reflect on the IMA website ........You list thirty species of Blennies.......and 79 fish tested within those thirty species.......thats less then three fish for each of the 30 species. Which means less then one blenny per species was included in the study per year {five years}Most of the fish species in your data have less then ten individual fish included in the data. How is it possible that so few fish can reflect accurately the industry? It may very well be that three fish in your data will be the sole representation for a species that is imported by the tens of thousands,from over one hundred separate islands and a thousand individual collectors. How can three fish reflect all the different collection conditions? if all three fish in your data came from one collector, that one collector would reflect all the collectors that collect that species!

God, it is just not sinking in, is it?

Peter already explained this one to you, Kalk.

Proper statistical analysis requires random sampling and a data set.
The more elements in the data set, the more statisitically significant the mean will be. If you want the mean to be reliable to within say 3%, you need sample sizes to be rather large... Let's say 1800 or so. Look at your stats book, look for the equation (P<=0.05). That says that the mean is 95% accurate. Read up on it- You can see what sample sizes are required to get whatever percentage accuracy you require, given that you know the population size too. This isn't rocket science- it really is just basic math.

Because the sample sizes were so low for some species, Peter lumped the species into families, then ran the numbers for families rather than species. You lose precision on a per species basis, but due to sample sizes, you get a much greater increase in confidence in the mean for the family instead.


So there- We've explained the same thing to you twice now.

CAN WE MOVE ON?

Are you really stating that three blue damsels collected during a five year time frame will reflect the other 800,000. Other blue damsels? If so then lets test three more, if the test is off by even one fish coming up different then the original three fish tested by Peter........the result will be a thirty percent change in the data. One fish! One fish can change the results by thirty percent on that species!

Third time. YOUR QUESTION HAS BEEN ANSWERED ALREADY!!!!

Go back and READ my post and Peter's post.

CAN WE MOVE ON NOW?

 

[mkirda]

the original three fish tested by Peter.

Hey, Peter!

When'd ya take the job with BFAR?
I didn't know you were testing fish now.

Regards.
Mike Kirda

 

[PeterIMA]

Mike, Did I say that?

Gun for hire,
Peter Rubec

Tran, I think he is talking about my CDT paper. The numbers are posted on IMA's web site. The main species he listed do not total 70% (see my ealier replies to Kalk). If you want a copy of the draft of my paper send me a PM.

Peter

 

[Kalkbreath]

Damsels alone represent around forty percent of the total fish exported from PI. this includes Clownfish , which are damsels. The data is supported every place that a break down has been tallied {you look it up yourself} If your too lazy to find out on your own then thats most likely why you cant comprehend the truth. Even a visit to any wholesaler of size will demonstrate that damsels out number any other species tenfold. The study that John Brandt listed on this board also lists the top twenty species imported for the hobby and PI does not ship some of the fish on the top twenty list.....like yellow tangs or Flame Angels. The Woods report also has the brake down I beleive and is a good read aswell. The Marine Conservation Society (MCS) is the UK Charity dedicated to the
protection of the marine environment and its wildlife. Please join us -
visit our website at www.mcsuk.org

 

[Kalkbreath]

I think this data should hush a few ......

A total of 1,471 species of fish are traded worldwide with
the best estimate of annual global trade ranging between
20 and 24 million individuals.Damselfish (Pomacentridae)
make up almost half of the trade,with species of
angelfish (Pomacanthidae),surgeonfish (Acanthuridae),
wrasses (Labridae),gobies (Gobiidae)and butterflyfish
(Chaetodontidae)accounting for approximately another
25-30 per cent.The most traded species are the blue-green
damselfish (Chromis viridis),the clown anemonefish
(Amphiprion ocellaris),the whitetail dascyllus (Dascyllus
aruanus),the sapphire devil (Chrysiptera cyanea)and the
threespot dascyllus (Dascyllus trimaculatus).The ten most
traded species account for about 36 per cent of all fish
traded for the years 1997 to 2002.Trade data,correlated
with aquarium suitability information,indicate that two
bluestreak cleaner wrasse (Labroides dimidiatus:GMAD
records 87,000 worldwide imports of this species from 1997
to 2002)and the mandarin fish (Synchiropus splendidus:
GMAD records 11,000 live individuals exported to the EU
in the same period).Data further indicate that species
characterized as ‘truly unsuitable ’,mainly due to their
restricted dietary requirements,such as the foureye
butterflyfish (Chaetodon capistratus),the harlequin filefish
(Oxymonacanthus longisrostris)and the Hawaiian cleaner
wrasse (Labroides phtirophagus),are also commonly
traded,albeit in lower numbers.

 

[Anonymous]

LOL, Kalk, your using GMAD numbers yet again.

 

[mkirda]

{you look it up yourself} If your too lazy to find out on your own then thats most likely why you cant comprehend the truth.

Kalk,

Laziness has nothing to do with it. You made a claim, I asked you for the citation. I refuse to try to look for a needle in a haystack- It would take me dozens of hours to find all the relevant papers, track them down and read them. Then I have to guess which one it was that you were referring to?
You think that is a useful expenditure of my time? I don't, and it is not due to laziness.

That knife cuts both ways.
Why are you so lazy that you still cannot give a proper citation to the paper?

I'm also still looking for those SD and SE figures for the set of numbers you posted. You aren't going to ask me to do the math for you too, are you?

Regards.
Mike Kirda

 

[hdtran]

Kalkbreath wrote:
{you look it up yourself} If your too lazy to find out on your own then thats most likely why you cant comprehend the truth.

Mike Kirda,

I think that particular comment (as opposed to other ones) was more likely addressed to me than to you.

Kalk,

Yes, I'm too lazy to find out on my own. I'm not an ecologist, and it's been over 20 years since I was active in biochemistry. I rely on other people to communicate to me their expertise. You haven't done so. Peter Rubec has. He has sent me a draft of a paper submission. The data in the draft appears to be self-consistent. (Same draft you have, I believe). If I am unable to understand your arguments, there are two major possibilities: (a) I'm too dumb to understand your arguments, (b) Your communications skills are poorly matched to written communications.

I looked at the IMA powerpoint, and found an aggregate %. You posted that (paraphrasing) 'more than likely 80-90% are clean.' Your argument for that is that one should only consider the most imported fish, and not all the fish tested. Fair enough, but that leaves the following questions:

• Which families are the most imported?
  What are their proportions?
  Were the samples taken randomly taken?
  Were the sample populations sufficiently large?

For 'randomly taken,' you have to trust the procedures. For 'sufficiently large' samples, I can draw my own conclusions (while I've never taken a formal college course in statistics, I know enough math to understand sampling).

Remember now, I'm an observer here. I'm not in the industry. Your mind is made up. Peter Rubec & Mike Kirda's minds are made up. Your arguing won't change their minds; their arguing won't change yours. I'm the undecided voter. So, either decide to take the effort to convince me, or decide to write me off.

You've told me the answer to 'most imported,' but you haven't given me the proportions. I want to see how you calculate that (paraphrasing) '80 to 90% of the imported fish are clean', and I want to be able to do that calculation myself, referring to either Peter Rubec's draft preprint, to the IMA powerpoint presentation, and a hand-held calculator. And tell me if you're using the numbers from Peter's preprint (which table) or from the IMA powerpoint slides.

Or else, conclude that it's not worth your effort to convince me, and I'll vote the Pat Buchanan chad 8)

 

[mkirda]

For 'randomly taken,' you have to trust the procedures. For 'sufficiently large' samples, I can draw my own conclusions (while I've never taken a formal college course in statistics, I know enough math to understand sampling).

'Sufficiently large' is so ambiguous- It is really about reducing the uncertainty or the error bars of the numbers.

Let's say, for example, we have a sample set of 10 vs. a sample set of 50.

If they both test out to the same percentage (mean), the Standard deviation and standard error will not be the same. If the mean is 20%, the error might be +- 20% on the smaller sample while it is only +-5% on the second sample. Again, assuming that the samples are reasonably random.

So a result of (20% plus or minus 20%) is not nearly as useful as (20% plus or minus 1%). This does not mean that 20% plus or minus 20% is WRONG. It just means that the error bars are such that it is more difficult to draw conclusions from the data set because of the smaller than ideal sample size. Ideally, scientists like to have the error bars be 5% or less. Not having access to the data set, only Peter can know what those bars are. I suspect that they are relatively high, hence the reason he lumped the fish into families instead of individual species. My own experiences with stats in college leaves me with a sample size of around 30 being about the minimum for having good numbers with reasonable error bars. My guess is that he doesn't have that each year for all species.
In this case though, in absence of any other source of numbers anywhere on the planet, you have to take what you can get.

And, it is not so much that I have 'made up my mind' so much as I understand the significance of SD and SE, and to what degree you can draw conclusions from the numbers based on them.

Looking at the Number of samples required, for 1000 fish species, you'd need 30 from each, minimum, each year, for the error bars to be reasonably small on every fish on a per species basis. That is 30,000 fish per year, or 120,000 over four years. Since he only has something in the 40,000 range, it is pretty clear that the error bars for at least some of the species will be too high to really draw comparisons if you stick to a per species basis. Cie la vie. It doesn't invalidate the results that can be shown. And this is the point that EVERYONE needs to get.

Regards.
Mike Kirda

 

[hdtran]

Mike,

I do understand the statistics, in spite of not having a degree in statistics. In my professional work, I do metrology, so some understanding of uncertainty, measurement uncertainty, and sampling is required. And please don't start talking Bayesian to me, or my eyes will definitely glaze over...

If I understand (and neither the thread nor the responders make it greatly easy to understand), the dispute has more to do with, 'are the aggregate numbers reasonably representative of the imported population.' Kalk is asserting no. He (I presume it's a he) asserts that one should only consider Pomacentridae, Pomacanthidae, and I-don't-rememberitae families. So, my question is, how does he arrive at the '80-90% of imported fish are clean.' Because I can't get 80-90% from reading either the IMA web powerpoints, or Peter Rubec's preprint, and Kalk has not written (or made available) a review showing how he arrives at the 80-90% clean figure.

I'm an amateur aquarist. I'm interested in sustainable practices, but I'm not a professional in this field. Just an interested observer.

Hy

 

[dizzy]

So, my question is, how does he arrive at the '80-90% of imported fish are clean.' Because I can't get 80-90% from reading either the IMA web powerpoints, or Peter Rubec's preprint, and Kalk has not written (or made available) a review showing how he arrives at the 80-90% clean figure.
I'm an amateur aquarist. I'm interested in sustainable practices, but I'm not a professional in this field. Just an interested observer.
Hy

Hy,
I don't speak for kalk (Jeff Koutnik). He's a dude all right.
Nor do I usually agree with him or want to be seen as agreeing with him. But I think I understand where the clean numbers are coming from. If most of the fish that were caught with cyanide die before they get sold, estimated to be as high as 95%, then it is logical that the ones being sold were most likely the clean ones. At least this is the way I interpreted what he was saying. It doesn't mean cyanide is not killing fish, or that it is sustainable, just that it is more of an economic issue for wholesalers and retailers than it is for hobbyists. I'm sure we'll be fortunate enough to have kalk come in and clear this up if I'm wrong.

PS
I'm not the one who uncloaked the kalkmeister. That honor falls to Jenn and Ad van Tage.

 

[PeterIMA]

Dizzy, I already replied to Kalk stating that I don't believe the idea that after 90% of the fish die through the chain that the remainder (those purchased from retailers by hobbyists) are necessarily cyanide-free (not cyanide caught). That seems to be a supposition on your part and some others. It might be true, but having watched fish die in my own home aquarium I doubt it. Fish being purchased by hobbyists most likely are still suffering from cyanide exposure. The fact that fish eat but waste away and die (ATP hypothesis) and other information about damage to internal organs leads me to believe they were caught with cyanide. If you don't believe me send fish to your local university and pay to have the fish tissues stained and examined histologically (sections then viewed under a microscope). At least one retailer in Gainesville FL did this and confirmed that most (about 80% for indonesian specimens) had liver damage (ergo, they most probably were caught with cyanide). There is no justification for claiming that the survivors at the retail level sold to hobbyists would all be the net-caught fish (some probably but not all).

Peter Rubec