Woodpecker ID Probabilities
From the Fallacy Files Weblog comes an intriguing probability puzzle involving a hypothetical population of red-headed and white-headed woodpeckers, and calculating the probability that a woodpecker that flies past an experienced birdwatcher and is identified as a white-headed woodpecker actually has a white head (i.e., is accurately identified). Given that the population consists of 85 percent red-headed and 15 percent white-headed woodpeckers, and that woodpeckers are correctly identified 80 percent of the time, what is the answer? Hint: it's much lower than you might expect! The author provides a detailed solution that delves deeply into the mathematical intricacies of Bayes’ Theorum.
This puzzle has direct relevance to the evaluation of sight records of Ivory-billed Woodpeckers in habitats in which Pileated Woodpeckers are known to be common. Given that there are so few (if any) Ivory-billed Woodpeckers in the Big Woods of Arkansas (or anywhere else, for that matter) relative to the number of Pileated Woodpeckers, and taking into consideration the "base rate" (an important principal that is often neglected, is explained here), it is more likely that a bird identified as an Ivory-billed Woodpecker is actually a misidentified Pileated Woodpecker.
Let’s assume, for the sake of argument and illustration, that a mature bottomland hardwood forest somewhere in the southern U.S. harbors two types of large woodpeckers, dark-winged and white-winged; 99 percent of the birds have dark wings and 1 percent have white wings. Let’s also assume that an experienced observer is able to correctly identify the woodpeckers to species based on wing color 99.5 percent of the time. Plugging these numbers into the formula, we find that the post-hoc probability of the observer being able to correctly identify a fly-by white-winged woodpecker is 67 percent (not 99.5 percent). The probability of a correct identification declines rapidly with declines in either observer ability or the proportion of white-winged woodpeckers. For example, if the observer’s ability to correctly identify flying woodpeckers to species is reduced slightly to a still-respectable 95 percent, then the post-observation probability of the observer’s identification of a white-winged woodpecker being correct plummets to just 16 percent! And if the proportion of white-winged woodpeckers in the population is reduced to 0.5 percent while holding observer ability steady at 99.5 percent, then the post-observation probability of a white-winged woodpecker ID being correct drops to 50 percent.
So, part of the problem of assessing the validity of sightings of Ivory-billed Woodpedckers lies in numbers which remain largely unknown: (1) the proportion of Ivory-billed Woodpeckers relative to Pileated Woodpeckers in various tracts of bottomland hardwood forests in the southern U.S., and (2) the probability with which field observers are able to make correct identifications of large woodpeckers. Given the current status of the Ivory-billed Woodpecker, the former number will always be at or near zero. The latter number will vary from one observer to another and from one habitat to another, and I have no way of knowing if it is close to 100 percent or nearer to 95 percent. Are the identification skills of volunteer observers tested before they are sent into the field?
At any rate, the base numbers alone seem to favor Pileated Woodpeckers to such a degree that "extraordinary evidence" would be required to accept any sighting of an Ivory-billed Woodpecker as scientifically valid.
This puzzle has direct relevance to the evaluation of sight records of Ivory-billed Woodpeckers in habitats in which Pileated Woodpeckers are known to be common. Given that there are so few (if any) Ivory-billed Woodpeckers in the Big Woods of Arkansas (or anywhere else, for that matter) relative to the number of Pileated Woodpeckers, and taking into consideration the "base rate" (an important principal that is often neglected, is explained here), it is more likely that a bird identified as an Ivory-billed Woodpecker is actually a misidentified Pileated Woodpecker.
Let’s assume, for the sake of argument and illustration, that a mature bottomland hardwood forest somewhere in the southern U.S. harbors two types of large woodpeckers, dark-winged and white-winged; 99 percent of the birds have dark wings and 1 percent have white wings. Let’s also assume that an experienced observer is able to correctly identify the woodpeckers to species based on wing color 99.5 percent of the time. Plugging these numbers into the formula, we find that the post-hoc probability of the observer being able to correctly identify a fly-by white-winged woodpecker is 67 percent (not 99.5 percent). The probability of a correct identification declines rapidly with declines in either observer ability or the proportion of white-winged woodpeckers. For example, if the observer’s ability to correctly identify flying woodpeckers to species is reduced slightly to a still-respectable 95 percent, then the post-observation probability of the observer’s identification of a white-winged woodpecker being correct plummets to just 16 percent! And if the proportion of white-winged woodpeckers in the population is reduced to 0.5 percent while holding observer ability steady at 99.5 percent, then the post-observation probability of a white-winged woodpecker ID being correct drops to 50 percent.
So, part of the problem of assessing the validity of sightings of Ivory-billed Woodpedckers lies in numbers which remain largely unknown: (1) the proportion of Ivory-billed Woodpeckers relative to Pileated Woodpeckers in various tracts of bottomland hardwood forests in the southern U.S., and (2) the probability with which field observers are able to make correct identifications of large woodpeckers. Given the current status of the Ivory-billed Woodpecker, the former number will always be at or near zero. The latter number will vary from one observer to another and from one habitat to another, and I have no way of knowing if it is close to 100 percent or nearer to 95 percent. Are the identification skills of volunteer observers tested before they are sent into the field?
At any rate, the base numbers alone seem to favor Pileated Woodpeckers to such a degree that "extraordinary evidence" would be required to accept any sighting of an Ivory-billed Woodpecker as scientifically valid.
posted by John L. Trapp at
8/25/2006 04:04:00 PM
45 Comments:
to at least some degree this is the sort of analysis where an old phrase comes to mind: "there are lies, damn lies, and statistics."
You are entitled to your own opinions, cyberthrush. But in taking the stance that you do on this, you ignore the long-established use ob Bayes' Theorem in helping to ferret out "false positives" in medical tests.
The very same principles are involved.
And this analysis really assumes the use of "robots" that have a set error rate. Human observers are not robots. If you throw in observer bias, you begin to understand that it's not statistics that are wrong.
It's the data, stupid! And how it is collected. This is why the purely "observational" studies that are published in many soft sciences like sociology or psychology can often be 100% wrong.
Observer bias can be devastating to a study. Going into a swamp where you believe a credible sighting of IBWO has occured raises your observer bias well beyond the error rate of your identification skills as a practiced field "robot".
"there are lies, damn lies, and statistics."
That has got to be the most tired, stupid, insepid, useless and misleading quote ever invented. And is usually only quoted by people who prefer anecdotal data.
Such people can often be heard to use the common English phrases of pseudo-intellectuals, such as "people say...", "I heard....", and "It is rumored that...".
It is ironic that the same people only ride in cars or airplanes and sleep in underwear designed for error rates of 1 in 6 million. I would think that such people would be happy to just jump off the airport's control tower and flap heartily to their next destination.
But no, they insist in cluttering up the airports with the rest of us. Oh well...
Either it is poorly written, or there is a serious error in this formula. It says, "The easiest way to see that this is correct is to suppose that the birdwatcher, while on a bird counting expedition, counts one-hundred woodpeckers in the forest. 85 of these birds will be red-headed and 15 will be white-headed." That's a faulty assumption. Though 15% of the birds are white-headed, that doesn't mean that 15 of the 100 birds you see will be white-headed. Let's go back to the 7th grade, when you first learned about probabilities. A coin has 2 sides. What is the probability that you can flip it 100 times and come up heads every time? Not very high - but it can happen. On the other hand, you aren't likely to get 50 heads and 50 tails, either. Same thing with the birds. You may see NO white-headed woodpeckers, you may see 1 or 2. And then there is another major flaw - a biological flaw. You are a huge variable in this puzzle. You move around the forest. Unless the birds have the exact same habitat preference and behavioral preferences, you are going to be moving in and out of best viewing sites for one or the other. Say that the red-headed just don't like to cross open areas...and are thus simply harder to see. You will likely see fewer red-headed birds, even though they are far more numerous. Even if you stood in one spot the entire time, you won't necessarily see 85 red and 15 white.
A comment on this topic submitted by Bill Pulliam of the Notes from soggy bottom blog was inadvertently linked to a different post. I have taken the liberty of reproducing it here in full. So, in Bill Pulliam's words:
Given the reported densities of Pileateds in the Big Woos, and the total population there is roughly on the order of 10,000. So add one Ivorybill to the mix, and you have .01% Ivorybills. The question becomes: Under what circumstances can the two species be distinguished with greater than 99.99% accuracy? Certainly a perched bird carefully studied with the head visible. Certainly not a brief flash of a bird tearing off through the woods and an impression of trailing white edges to the wings. But where is the line in between? I don't think this will ever realistically be knowable. But in a situation like that (0.01% Ivorybills), the odds of being able to unequivocally document that one bird are vanishingly small. This species will in all likelyhood never be thoroughly proven to still exist unless someone finds a pair tied to a small home range by a nest. There, the ratio becomes more like 100:1 and the probabilities of encounter and 99.9999% certain IDs increase 100-fold.
And if no one ever does find this situation, well, it's all moot anyway, isn't it? With no reproduction you don't really have a species anymore.
Getting my comment on the right post this time( oops!)
Mr/Ms Anonymous #3's arguments are not actually relevent to a probability model. You could add the phrase "on average" in front of each number, and the conclusions are unchanged; it's just more tedious to read. That observer will ON AVERAGE see 15 real white-headed woodpeckers and ON AVERAGE 85 red-headed ones. S/he will identify ON AVERAGE 68 redheads as redheads and ON AVERAGE 17 redheads as whiteheads. S/he will identify ON AVERAGE 12 whiteheads as whiteheads and ON AVERAGE 4 whiteheads as redheads. So, ON AVERAGE s/he has identified 72 birds as redheads, 5.6% of which are misidentified. ON AVERAGE, s/he has identified 29 birds as whiteheads, 58.6% of which are misidentified. This is the way probabilistic events are always examined; it is perfectly valid and time-tested.
It is also worth noting that ON AVERAGE our hypothetical birder will identify 29% of the birds as whiteheads, which is nearly double the true abundance of the species. And the error rate affects the rare species far more than the common one. This is why we count a fleeting glimpse of a Northern Cardinal in Tennessee, but we don't count the same fleeting glimpse of a Northern Cardinal in British Columbia.
As anyone who has studied statistics learns right up front, statistics NEVER allow you to conclude the actual pure boolean truth or falsity of a single observation. It just lets you estimate the probability. "You cannot make inferences about an individual from group statistics" is one of the fundamental axioms of the the science of Statistics; and one of the ones most often violated by non-statisticians: On average men score better on math tests that do women, therefore this man will score better than that woman FALLACY OF THE PRIME ORDER!
ALL data, even in the "hard" sciences, are a matter of probability, not god-like absolute certainty. And ALL data are "observational." God does not reveal things to us, we have to go out and measure them: i.e. MAKE AN OBSERVATION. Whether you use your eyes and some Mall*Wart binoculars or a $4.5 gazilion piece of equipment and a supercomputer, it is still an OBSERVATION. Science does not deal in absolute truth. Slandering the "soft" sciences just because their data are more error-prone is, well, unscientific.
This is relevent FAR beyond big, crested, swamp woodpeckers. Most of us who are long-time birders of the sort that actually do report and submit our rare findings for peer review have acquired a sort of intuitive understanding of this phenonemon: an extremely rare sighting might more probably be an extremely unlikely mis-ID than an extremely unlikely bird. It is a bit humbling sometimes to put numbers to things, though. All the arguments about observer bias etc. don't really undermine the basic model; they are simply some of factors that affect what the error rates are. It also doesn't matter whether the error rate is caused by observer error or by 20% of the birds actually having the wrong color head.
Now I have to say this example uses an enormous error rate of 20%. In reality I'd think our actual accuracy rates are more like 99-point-something percent, for reasonable experienced and cautious observers. I mean, really! Anyone who can't tell Red-headed Woodpeckers (Melanerpes erythrocephalus) and White-headed Woodpeckers (Picoides albolarvatus) apart with an accuracy better than 80%, even on a quick flyby, should have their binoculars confiscated!** But the actual value of that "something" in the "99-point-something" matters enormously and is all but impossible to determine most of the time.
**This is a joke; I suspect the person who wrote the example is not a birder and isn't even aware that actually are two real species with these English names!
Forgive me for following up on myself, but I've been pondering this whole "error rate estimation" problem over breakfast (after watching a Hooded Warbler chasing a Blue-winged Warbler around the pond.. entirely unrelated). I think I can take a stab at a very rough ball-park estimate of the base rate of PIWO/IBWO mis-IDs. In a medical test, you establish the false positive rate by running a bunch of known negatives and looking at the results. We can do sort of the same thing here.
I am an experienced decades-long southern birder. There are hundreds like me. The large majority of us have never identified a single bird as an IBWO (correctly or incorrectly) in our entire birding lives; this includes me. Not a kent, not a glimpse, not a raprap. So, how many Pileateds have we encountered without a single positive (false or otherwise) IBWO ID? Allow me the back of a very sloppy envelope here:
Let's say 100 PIWO encounters per year for our hypothetical typical experienced southern birder. This is a couple per week. Some of us who live in good habitat might see them more; some who live in more urban or agricultural areas will see them less. Just an order-of-magnitude here. So, this is about 1000 PIWO encounters per decade. Our typical birder IDs far less than one IBWO per decade, so we have an upper limit of the error rate at 0.1% But in fact, TEN of these typical birders don't ID a single IBWO per decade between the lot of them. So the error estimate is pushed down to roughly 0.01% or below. In my lifetime, I have only known one experienced birder who claimed an IBWO encounter (over 30 years ago). And, though he had great knowledge, taught me and many other novices an enormous amount, and found many legitimate and verifiable rarities.. frankly he also had a disturbing tendency towards "stringiness" as the Brits call it and his uncorroborated rarities were not always held in the highest regard. So let's count that as a false positive. I've probably known somewhere between 10 and 100 of these "typical southern birders" well enough that I would know if they believed they had ever seen an IBWO. So the scribbles on the back of my envelope give a very rough estimate of the false positive rate for IBWOs amongst these birders as between 0.01% and 0.001%; i.e. one IBWO booboo for every 10,000 to 100,000 PIWO encounters. Personally I would suspect the low end of the rate, closer, to 0.001%, since I really have known more like 100 of these folks and the one IBWO comes from the most suspect source.
This puts the base error rate in the realm where one IBWO among 10,000 PIWOs in the Big Woods might be barely detectable, with real encounters possibly outnumbering false positives. But the signal-to-noise ratio is not going to be very good. And you do need to keep in mind that in that habitat the PIWO abundance is the highest I have ever seen anywhere, and the rate of PIWO encounters is more like 10-20 PER DAY, not per month. Thus it would take about 10 volunteers out every day for 100 days to see all 10,000 of them, and to give you a shot at seeing that one real IBWO (if it is there) one time. Cornell's search this last winter achieved about this order-of-magnitude level of intensity, but only just.
OK I'll shut up now.
Thanks for the elaboration, Bill. Very interesting "seat-of-the-pants" calculations!
statistics applied to inaninmate objects and roullette wheels is one thing, applied to human behavior is quite another -- the original analysis above is so fraught with oversimplicity (leaving out 99% of the variables) it's just meaningless to apply it to any single instance -- it's fine as a mental exercise I suppose but it says absolutely nothing about whether or not Bobby Harrison saw an Ivory-bill when he says he did. If you flip a fair coin enough times than statistically at some point you'll get a run of 100 heads, and if you put enough monkeys on enough typewriters for enough time, one of them will type MacBeth -- duhhh, SO WHAT!! And human behavior is infinitely more complex than this. I LOVE mathematics, but statistics and the way their applied to behavioral science is another matter.
Probability theory and statistics are among the basic tools of all natural science. They are what allows science to take place in the face of indeterminism and uncertainty.
Bill's error rates are totally too small. Just look at what happened when Cornell started insisting that 2 birders go on every survey. The error rate went way down! That is less bogus IBWO were seen.
Cornell knows it. They fixed the problem with that little fix they made.
No, observer bias was way large when Gallagher went into that swamp. He wanted desperately to see. And he did!
It's just that he was in error.
bill, bill, bill.
The soft sciences are not "soft" because their data are more error prone. It's because their researchers are more error prone. And their journals allow more anecdotal evidence.
I seem to be catchin' it from all directions at once here...
Anomymous #4 says: "Bill's error rates are totally too small"
Care to elaborate why? It seems to me the fact that large numbers of experienced birders in a region where Pileateds are common virtually never report Ivorybills is pretty good evidence that regional birders are highly skilled at not making this mistake. Do you have an alternative analysis, or just an opinion?
About the group-think, expectation bias, whatever thing:
You seem to have forgotten that the first Arkansas followup sighting was by two observers simultaneously, not a solo.
With one exception, in all the other cases that I can think of where an Ivorybill has been reported since the 40s, the followups by experienced birders (not counting the initial sighter) have resulted in zero additional sightings. This has been true even when the followers-up were hopeful, optimistic, and going in with an expectation of encounters, such as the "Pearl Epedition;" the only followup sighting there was not by an experienced birder. This argues rather strongly against the theory that if you send skilled birders into the woods expecting Ivorybills, they will find Ivorybills even if there aren't any there.
The one exception to this pattern that I know of is Aggey-Heinzman (probably spelled wrong). In this case, followup lead to additional sightings as well as the only physical evidence of the species that has been collected in the last 60+ years. This raises the significant possibility that this time the followup sightings happened because the birds were indeed actually present.
So there's not much pre-Big Woods precedent for a flurry of spurious Ivorybill sightings by experienced birders in the wake of an initial report. These are not Selasphori or Empinonaces. They are two boldly-patterned woodpeckers in different genera.
"The soft sciences are not "soft" because their data are more error prone. It's because their researchers are more error prone."
ridiculous, the "soft" behavioral and biological sciences are FAAAAR more error prone than the physical sciences, simply because they cannot take into account all the pertinent variables -- they toss statistics onto their weak data to give a false sense of scientific precision that doesn't exist. Even in a field like molecular biology much of what is believed today will be vastly altered or even overthrown 100 years from now, because of false assumptions made. On the other hand, an IBWO was seen in Arkansas and nobody has proven otherwise, just invented an alternative hypothesis, which again one can do for any scenario.
Bill, bill, bill.
Your error rates are way to small. You proved my case. In every case you cite from your experience, the observers didn't go into the woods expecting a chance to see IBWO. Therefor, they didn't see one. ( and, of course, due to the fact that they ain't there)
You marginalize the observer expectation bias enormously. Why, this whole fiasco is on enormous observer expectation fiasco. The only people who ever have seen IBWO in the swamaps are those who "from childhood" dreamed of finding one. Harrison, Gallagher, Kullivan. Huge observer bias!
Then look what happens. Exactly what you would expect. The follow-up sightings taper off as reality reduces observer expectation bias.
It's classic. This whole fiasco will be a great lesson in my statistics classes.
Cyberthrush, your world-view is so hopelessly skewed that I don't think you could even understand the intricacies of what we are talking about. Everything you say, says that the soft sciences have poor researchers who don't know how to take data. Just read what you wrote. You REALLY proved my point!
I think one thing is definitely true.
Cornell thought they had a case of Observer Expectation Bias. That's why they went to two person teams last fall. They were just having too many I-saw-the-white sightings without photo documentation.
It was a courageous admission and adjustment. It's what scientists do. I applaud them for that.
Bill, bill, bill.
Btw, I agreed with everything you wrote about your statistics in the absence of observer bias. It's just that I am arguing that your statistics assume that your system that is being observed is stochastic and non-time varying.
That last, non-time varying, is the biggest quandary for good data.
Your data error rates on you and your friends and acquantances may be right. But they assume a condition in which it was folly to report IBWO and thus had a low average expectation rate.
When Gallagher went into the woods. The Average expectation rate had risen. It vary in time. Now your statistical rate is too low. He entered the woods with a much higher expectation rate. A true observer bias that could not be controlled for by a good double-blind test or other control methods.
And I agree, Cornell could only control for Observer bias by using two teams.
Well, guess what? It worked. They didn't find any IBWOs.
Btw, Mr. Trapp, excellent post. You actually added to a much discussed subject.
Several other people than just those you list reported follow-up sightings in Arkansas, including Melanie Driscoll. She reports her reaction to her own sighting was disbelief, because it was impossible, because the bird was "freaking extinct" (her own words). Doesn't sound like bias from observer expectations to me. I'm not saying the phenomenon does not exist. It's a very well known problem with birding, with many infamous cases of it. It is also something we TRY to compensate for. Most of the notorious cases of massive observer delusion center around species that are far more easily confused, even with a good look, than D. lineatus and C. principalis; although admitedly the Smith's Longspur / Common Skylark thing back in the 70's was amusing, it was also sorted out quickly.
I don't know if you are a birder, but if you are not, you might not understand the HUGE wall of taboo surrounding an Ivorybill sighting. This is not something that anyone with any concern for their credibility claims lightly. This taboo is a very strong counter force against the elevated expectations effect. There is no similar taboo against reporting a Skylark in California or a Rufous Hummingbird in Illinois, or even a Little Stint in Kentucky. The only other two species that are even remotely similar in this regard are Bachman's Warbler and Eskimo Curlew. And, as I pointed out before, in past times even when a seemingly credible Ivorybill sighting has been reported, the followup has NOT lead to a rash of additional spurious sightingsby exerienced observers. The waves of spurious sightings come from the hunting/fishing/boating/backyard feeder watcher crowd (by the thousands!!!), not from the serious birders.
Evidence, I give you evidence. The expectation effect is a good theory; there's just not much evidence that it actually happens in the case of experienced birders and the Ivorybill (unless you want to engage in the circular reasoning of "it happened in Arkansas which proves it happened in Arkansas." Now whether these sightings individually or collectively are sufficient to fully document the occurence is a whole 'nuther matter. I'm just adressing your claim that they are "bogus" and just "observer bias."
Thank you for the compliment. I'm glad that I could add something positive to the discussion.
The point about time-varying observer bias is quite valid. I'm not convinced, however, that expectations were any higher in 2004 in Arkansas before the first followup sightings than they were in 2000 in the Pearl, and I am still struck by the contrast between zero followup sightings in the 2000 case versus a half dozen followup sightings in the 2004 case. I know for myself, the failure of the 2000 search to find anything actually lowered my expectations. I was far more hopeful about the Ivorybill's survival in 1999 (and 1989, and 1979) than I was in March 2005. But these psychological variables are ultimately unknowable and highly individual.
Another thing that can easily vary over time is the actual presence of a bird. Birds move. Frequently. Often long distances.
Observer bias and all, I still find it impossible to synthesize the best of the followup sightings (such as Driscoll's and Taylor's) from a Pileated, without invoking outright delusion, hallucination, or fabrication. These were not fractional-second glimpses of white that might or might not have really been there.
Bill, bill, bill.
I don't know if you are a birder, but if you are not, you might not understand the HUGE wall of taboo surrounding an Ivorybill sighting
You are making my point. What you mean to say is that their WAS a huge wall of taboo...." But that wall fell before and after Mr. Gallagher. He wanted to believe. His friend had seen and it was his life's abition.
The average expectation rate changed. It time varied to a new distribution rate. He was primed with Observer Expectation.
I am a birder and a statistician. How can you possibly say Observer expectation bias doesn't exist. It is famous in birding. 100's have mis-id'ed a bird for days that was first incorrectly identified. I've seen it personally many times in small examples of birding with friends. If only momentarily.
Remember, Bill, we are talking about a huge Taboo that moved to a "possibility". A huge time variance in the stochastic probablities. It's an enormous change. From near zero to possible. Wow! How could someone not see the bird?
"How can you possibly say Observer expectation bias doesn't exist"
Excuse me while I spit some words out of my mouth that I didn't put there. *PTOOOEY*
I in fact said (in the very same post you quoted, so I know it wasn't a case of comments passing in the 'net):
"I'm not saying the phenomenon does not exist. It's a very well known problem with birding, with many infamous cases of it. It is also something we TRY to compensate for. Most of the notorious cases of massive observer delusion center around species that are far more easily confused, even with a good look, than D. lineatus and C. principalis"
These circles are making me dizzy.
The expectation effect is a good theory; there's just not much evidence that it actually happens in the case of experienced birders and the Ivorybill
Bill, bill, bill.
Put that "ptooey" back in your mouth and read your own words.
Observer expectation bias is real and wide ranging and is all over this IBWO fiasco. Any good scientist knows it.
It's not circular. It's a fact.
So to sum up. Here is the devastating effect of the Observer Bias argument.
1.people are not robotic automatons that have set error rates during observations.
2.instead their rate of error is influenced by human factors.
3.they change with time due to wishes and wants and their own expectations.
4.these expectations, wants, and wishes are admitted to in various books, lectures, letters and published articles. They “grew up” wishing to see the bird, “dreamed” of seeing it, searched many years, talked to like minded people who shared their desires, and they wish to be recognized in their field of ornithology or birding. (ambition is not bad, just it’s influences)
5.No double blind studies of IBWO sightings is possible.
6.To minimize observer bias, multiple persons per team would be expected to reduce not increase sightings if the bird is extinct.
7.To minimize observer bias, impartial observers would be a plus. But nearly impossible to get all impartial observers on a team. What would that even mean?
8.Observer Bias predicts all that we have seen. Initial sightings, followed by declining sightings. Ultimately producing no real documentation or proof.
OK so I'm a bad scientist now. Enough.
Explain why this pattern you predict did not happen in 2000 in the Pearl River expedition. There were no sightings by any experienced observers, and I believe only one in total (from a non-birding psychic-consulter). Yet your model predicts there should have been multiples. In that case there was national media attention at the time of the expedition -- even higher pressure than in Arkansas before the public announcement. If you need evidence that there was strong desire-to-find, look no farther than the ARU recording of obvious gunshots that they tried to force to be woodpecker raps. Yet even in the face of this push and bias, there were still no spurious Ivorybill sighings generated. This is a test of your model's ability to predict past results, and in this case it fails to predict accurately, even a posteriori.
Reread Casey Taylor's full account of her sighting (a bird in flight watched through binoculars for several seconds, with clear description of both underwing and upperwing patterns) and explain how that is generated from a Pileated purely by observer bias unless she is a terribly incompetent and sloppy observer. I refer to her full account published in North American Birds, not the small synopsis in the Science paper. As a side note, in my opinion anyone out there who has not yet read the North American Birds paper (Rosenberg et al. 2005 59(2): 198-206) should do so before discussing the value of the sightings.
There is evidence that there was more going on in Arkansas than just observer bias and wishful thinking. Conclusive? Obviously not or this debate would be long over. But nonexistent? No.
By the way, the circles I refered to were not your reasoning, but the stagnant back-and-forth between you and me (whoever you are).
Bill, bill, bill.
I would never call you a bad scientist. I said "any good scientist" like you. I was impressed with your reasoning. I just think your conclusions are wrong because of a missing factor. Ok, I admit I left off the "like you" part.
No, "my model" does not predict any such sightings. What is the variance in the error distribution for Gallagher, Kulivan, et al? No one knows. The tapering off can be very fast. Indeed, Kulivan was the first to be recently "plausible" sighting after many years of "huge Taboo" as you say. Those that followed Kulivan into the swamp probably still bought into the whole impossibly "Elvis" Taboo.
The subsequent searchers had much lower expectation bias than did than did those you entered the Big Woods after Gallagher. So the taper was much faster after Kullivan.
Fits the model beautifully. But it's not my model. It's a well-known statistical bias.
It's the soft sciences curse.
Reread Casey Taylor's full account of her sighting (a bird in flight watched through binoculars for several seconds, with clear description of both underwing and upperwing patterns) and explain how that is generated
Bill, bill, bill.
Can't you give me a tougher challenge? This is so easy. A few seconds glimpse thru binocs? Come on.
Look at the CLO. After a year and more at looking at the Luneau video of black and white, they still can't sort it out correctly. Why? I not sure in the beginning. They just aren't as good a Sibley, probably.
But now. Now that they have committed to there course, they will never admit. It's a case of extreme Observer Bias. Everything they see in the blurry video is IBWO. Isn't that amazing?! All their professionalism is gone.
Sad really.
"There is evidence that there was more going on in Arkansas than just observer bias and wishful thinking. Conclusive? Obviously not or this debate would be long over. But nonexistent? No."
Bill, bill, bill.
Well, which is it? You can't have it both ways. The bird either existed in the Big Woods or not. First you say it's not conclusive. Then you say it's not nonexistent.
Well, which is it? It's one or the other.
If you allow yourself to base your predictions on infinitely-adjustable suppositions of the internal mental states of individual people, I suppose you can make anything fit the model. Pearl searchers: low expectations ergo no sightings; Arkansas searchers: high expectations ergo multiple sighings. These assumptions about contrasting expectations in the two situations are arbitrary and post hoc, tuned to force agreement between prediction and observation. You are creating a special case of assumptions for each incident. As I said earlier, an argument can be made just as strongly that the failure to successfully confirm the Kullivan sighting in spite of a large-scale effort actually raised the bar and lowered expectations, rather than vice versa. I don't think the Taboo was really lifted until after the series of follow-up sightings in Arkansas had already begun (since that is what lifted it, after all).
But now there I go, making suppositions about the internal mental states of individual people.
It seems observer bias affects the interpretation of observer bias. Ah well. I've always said that ultimately all human reasoning is circular, the trick is just to make the circle big enough that it looks like a straight line.
Hip-deep in the irony, I think we have reached the point in this discussion where we are likely to remain, yes?
If a breeding population still exists, someone will likely find it in the next few years and we will actually have incontrovertible evidence (I think "extraordinary" is a bit too hyperbolic of a word; I think "clear and unambiguous" is sufficient). Enough people are looking now that the "nobody has been searching for them, that's why they've been missed" argument will evaporate steadily over the rest of this decade. And if such a population doesn't exist, it's too late already and it is moot what Casey Taylor or David Luneau's video camera actually saw.
I agree, Bill, that "extraordinary" evidence is perhaps a bit too hyperbolic. At this point, I would be pleased to see either incontrovertible or "clear and unambiguous" evidence.
Bill, bill, bill. John, john, john.
Perhaps, extraordinary refers more to the fact that finding a live IBWO would in itself be so extraordinary that proof of it would in fact itself be extraordinary.
So in effect, a simple non-blurry photo of an extinct IBWO would be so extraordinary as to blow people's minds.
As for our other aruguments, we are indeed now circular. I can only hope and pray that research scientists such as yourself take bias into account. Or else, certain sciences will continue to have a bad name. And certain institutions also.
Bill, bill, bill.
But on the other hand, Bill, you know I didn't just infinitely adjust my expectations. I only adjusted them once on my observations.
Clearly, Kulivan's searchers were much less ready to accept that your "huge Taboo" had fallen. Than were the Gallagher searchers who for the most part were volunteer birders from all over the country who were already true believers.
And why wouldn't they be? A premier bird research center had told them that it existed and it was only a matter of time.
Tell me that isn't High Observer Expectation Bias!
"it is moot what Casey Taylor or David Luneau's video camera actually saw."
Bill, bill, bill.
No it's not moot. It's a fabulous, fabulous lesson for statistics classes. And a warning to Journals and would be science authors.
It's a modern day bigfoot taken seriously by serious scientists. That example doesn't offer itself every day.
Just a point of fact:
searchers who for the most part were volunteer birders from all over the country who were already true believers
These 100 volunteers from all over the country failed to produce anything but one "level one" (the lowest level) sighting. The earlier cluster of sightings before the public announcement were made by a magazine editor, an art professor, a nature center director, and five CLO employees. Two of these sightings were made by teams of two observers simultaneously, the remainder were by solo observers. Maybe the "true believer volunteers" were better birders who made fewer mistakes, maybe the bird responsible for triggering the sightings had vacated the area by the time they arrived. Regardless, it is good to keep the facts and chronology straight, and it was not these volunteers who produced the cluster of sightings.
"These 100 volunteers from all over the country failed to produce anything but one "level one" (the lowest level) sighting"
Bill, bill, bill.
Your chronology stinks. You are confusing the last year of search and the first year of search. Now remember, Bill, the CLO adjusted for Observer Expectancy Bias by making the second year teams go in pairs.
That greatly reduced the sightings just as CLO expected it to do. Or at least, worried that it might.
It worked. The scientists did what scientists do. They removed bias from their surveys. The frenzied High Expectation Birders could no longer "see" by themselves. They were corrected by their partner.
I bow down before CLO for making the change. Even they know about Observer Expectancy Bias.
Huh???????????
I was just addressing your reference to "volunteer birders from all over the country," and pointing out that these were not the ones who reported seeing Ivorybills. Most of the post-Gallagher/Harrison sightings were made by CLO employees from New York, plus a few special guests. There were no volunteers from all over the country during the first search year, when the sightings occured, which is I thought the time frame we were talking about?
I give up. This is pointless. I'm tired of being patronized and told I'm an idiot by someone who hides behind a smokescreen of anonymity. Enjoy.
The probability that people will argue about whether the IBWO is extant even if someone brings in a live bird in a cage (can't shoot one, and even if you could, someone would claim that it was an old specimen in unusually well-preserved condition): 100%. Probablity that Trapp will ever again write an entry that invites discussion (?) of IBWO? I'm guessing ZERO!
Don't be so sure about that. I'm a slow learning and somewhat of a glutton for punishment! Besides, this has been sort of fun, even though I did start to get a little dizzy reading all of the back-and-forth comments.
What is this bogus reasoning that Skeptics will never accept a valid sighting of an IBWO? It's nonsense. It's just believers still trusting in crappy evidence. Just mad because we insist our birds to be real and properly evidenced?
.....also, the difference between a blog with comments and a blog with none or few is like day and night, spring and winter, good and evil, flowers and death...
...oops, sorry, I got carried away...
Like Birds Etc versus Ivory Bills LiVE. All the difference in the world.
I guess some people that have a hard time letting go, and refuse to let reality alter their beliefs.
Interesting in light of some of these comments about observer bias is that, as noted in Gallagher's book, in the first field season none of the most experienced and skilled observers from Cornell ever reported any visual or aural evidence of IBWO despite being the first to begin searching after the initial reports.
Post a Comment
Subscribe to Post Comments [Atom]
<< Home