Alright, it's been a while. Where were we? Oh yes...
train;289999 said:
(Statistical) Analysis is one of two things:
Complete - with all components intact, meaning there are no variables and no assumptions, the facts are the end all...
Incomplete - with some components intact, meaning there are variables and assumptions that must be made in order to attempt arriving at a feasible hypothesis.
(example at the end of all this to also address the statistic question...)
Okay, I don't know where you got this, but your first category looks like a myth. Something that doesn't actually exist. Or rather, it could exist in some sort of simple system, but then it wouldn't be statistical analysis at all. It wouldn't be statistical for sure and I don't see the point in calling it analysis either. For the sake of the argument, assuming something has ever fallen into your first category and also deserves to be called statistical analysis, I contend that your second category does not encompass all of the rest of statistical analysis. In particular, statistical analysis is not always done with the goal of generating an hypothesis. Actually, I can't think of any examples where that is the goal, although I could conceive of unusual cases where it would be.
If solar Variability were not as much a factor, then there wouldn't be much difference between records as shown
here
What? Of course there are regional differences in received solar radiation. I would think that should be obvious. It also has nothing to do with this. I thought we'd already been over this. Global glacial cycles, anthropogenic global warming, etc. are distinct from (typically more obvious) regional variation in things like cloud cover.
Because ice cores have been taken from geographically disparate regions, it is unreasonable to conclude that regional variation could have a systematic confounding effect. And if the effect isn't systematic, it can be detected through discrepancies in cores from different regions. I know I've said something like this before...
...all of which is almost moot when it comes to this particular point: as far as I know, solar radiation could not affect isotope fractionation outside its effect on temperature, so the temperature readings aren't any more or less accurate from this.
As for Volcanic activity - the ash particles that are brought back to earth easily by precipitation do not provide the "global" records. The particles providing the "global" similarities between cores for volcanic activity are the ones sent into the upper atmosphere that have the potential to travel farther from the source. So one eruption can taint multiple core samples - but does not mean that the climate model based on those samples is accurate.
What? This doesn't even make sense. I thought that earlier you were contending that ice core data could be confounded locally by volcanic eruptions. Now you're talking about a global effect. If it's global, then your hypothetical problem goes away anyway.
It wasn't an assumption. I stated it because they found a global source for similar sampling, but not all the flow modeling is similar. A correlation would be to dip a beaker into the Central Pacific Ocean and another the North Atlantic Ocean, and not take into account any ocean currents or tides. "How many similarities can I find. Hmmm... well - this must be how things were at the time."
That isn't what the word "correlation" means. I gather that you mean "analogy."
And no, that isn't analogous. It's completely different on so many levels that...actually, I can only think of one level on which it's at all similar. There is a whole lot of ocean. And there's a whole lot of atmosphere too. And both for your hypothetical beaker-dipping and for ice cores, only relatively small samples are being taken. But there the similarity ends. Every single other thing about ice cores is completely different from your example.
Perhaps most importantly, ice cores exploit a naturally occurring record of the past. Your beaker example doesn't.
But actually, if we set all that aside, the one similarity between them has a key difference that perhaps you're overlooking. Your beakers are a very small sample. And I don't mean relatively small. I mean small. It isn't the size of the population or the size of the sample relative to the population that's important for statistical analysis. It's the raw size of the sample. Many ice cores have been processed and more data from ice cores is being collected right now. The sample continues to grow. To be fair, you should be dipping dozens of beakers into several ocean locations. And actually, you could get some good data from that. Let's say you were interested in the concentration of a certain chemical in the ocean. If all of your beakers were coming up with similar numbers, you wouldn't need to worry about currents or tides. They don't have any means of affecting the results. If I objected that you hadn't factored them in, you could rightly disregard my objection.
In paternity tests, you have both the input and the output so the analysis is complete.
Wrong! Sequencing an entire genome is expensive. Oh, it's been done. And it's getting cheaper. But there's no reason to do it for a paternity test, as "incomplete" methods work just as well.
If they don't equal - it ain't happening. In other DNA tests you usually have two outputs - then there are assumptions to be made and analysis is incomplete.
Yeah, I don't know what got you so hung up on this concept of "completeness." It doesn't reflect at all on how scientists actually work. It's almost always impossible or impractical to have all of the data on anything. Sampling works.
statistical techniques are meant to provide analysis in unbiased manners and they can be put to use, when the analyzers also maintain an unbiased stance... However, statistics are methods and means that man has created in order to best represent what they are trying to communicate/interpret. But there is a difference in the type of statistics being used with ice cores... I stand firm on Ice Cores being incomplete.
It's fine if you don't understand how ice core analysis works or how DNA testing works. There are several components of both that I know little to nothing about myself. These are complicated things.
You also seem to have little understanding of statistical techniques. That's fine too. Just don't preach about them then.
Example of Complete: 3 variables, all accounted for
A Class of 100 students were surveyed on choosing all(X), some(Y) or none(Z) of something. All responded.
- 37% chose X
- 28% chose Y
- 35% chose Z
Determinations - all definite.
That's actually only one variable (with three levels). See, this is what I mean about the preaching. I don't mean that to be condescending and maybe another term would be better, but that's how this comes across to me. At any rate, you are trying to explain statistics to me, but you're getting something wrong that is so basic, it signals to me that you don't actually know anything about statistics.
Again, I don't want this to be harsh. But I can't really come to any other conclusion when you go and do something like this.
Example of Incomplete: Only 3 variables, some unaccounted for, with a relatively small margin of error... but major determinations from such a small margin...
A Class of 100 students were surveyed on choosing all(X), some(Y) or none(Z) of something. 92% responded. 8% did not respond, or surveys were lost, etc.
- 32% chose X
- 26% chose Y
- 34% chose Z
So...
According to another survey, the remaining 8% would have been 5 for X, 2 for Y, and 1 for Z.
Or
Logically, based on current rates, with the remaining surveys, Z would have had the most, then X, then Y
Or
IF Y was as appealing as Z, it could sweep remaining surveys and tie Z based on preference of Y over Z and X would have the least
Or
IF X was most appealing, it could seep remaining surveys and have the most, followed by Z, then Y
Or
Z could have the most, while X and Y tie
etc...
So - the more variables, which is what the ice cores are taking into account, the greater the mis-determination possible. When you begin to increase the possibility of mis-determination, you continue to decrease the possibility of certainty.
Not to beat a dead horse, but this is also wrong. Yes, ice cores generate data for several variables, but that's because those are the things scientists found meaningful and figured out how to measure. An ice core doesn't say to itself, "I have ten variables." Scientists just measure what they can. If, for example, they later go on to compare the data for carbon dioxide concentration (based on direct measurement) and temperature (based on isotopic fractionation) the data for those two variables isn't somehow less accurate just because they also happened to measure several other variables from the same core.
Ice cores are still soooooo young in the process, that when compared to some of the recent dinosaur fossil determinations, you wonder just how off the mark they may be. (IMO)
Yes, I do wonder! But like I talked about earlier in this thread, with practical applications for the same techniques that are being used in ice core analysis, it would seem that they can't be
that far off the mark.
Not sure what you mean about dinosaur fossils. Care to elaborate?
EricBess;290010 said:
This actually brings up another question. Again, I don't know much about the research, but does the posibility exist that for some sort of distribution of temperature? I mean, isn't it possible that when the temperature over the ice cores is warmer for an extended period of time that the temperature over other areas is necessarily cooler during that same time period, rather than an assumption that it must be warmer also?
But we would have seen that effect unless it magically disappeared some time in the 19th century. Also, there's no mechanism for it. If it's just Antarctica we're talking about, then sure. But now that we've got ice cores from other parts of the world, in order for the temperature to be higher over all of those places while it's simultaneously lower than normal over the warmer places in the world would require something to actually cause that. Not necessarily a heat-shifting sky-genie or anything (that's the first thing that came to mind), but still it would have to be something unprecedented in scientific discovery thus far.
I also agree with train that the ice core science is still incredibly young and we need to be careful making too many inferences.
Young like genetics?
Oversoul - I don't see how extending this to a paternity test is even remotely related. A child has 2 inputs (for lack of a better analogy) and 1 output based on those inputs. A paternity test takes information from the original output and an unknown to give a probability that the unknown was likely one of the original inputs. There isn't much margin for error here. The possibility exists for a false positive, but there is sufficient variety for this to be unlikely. A false negative would be nearly impossible. There are far more variables in ice cores. Perhaps I have misunderstood what you were trying to say.
Okay, so when they do a paternity test, they don't generate complete genomic sequences because it's time-consuming and expensive to do so. What they do is clone the DNA a bunch, then use special enzymes to extract DNA from certain regions of the genome containing repetitive sequences. A lot of the genome is repetitive, actually, but the parts that are of interest in paternity tests are highly variable among the general population. Get enough of them to line up, and it's astronomically unlikely that the similarity is due to chance.
For a variety of reasons, a single ice core isn't likely to have an unbroken record of atmospheric data. I was saying that to reject ice cores on this basis would mean that one should also reject paternity tests and most other DNA tests. They don't have unbroken records either.