Name the Major Flaw In This Signal Processing Analysis Probl

[Clay]

On Dec 13, 5:16 pm, Rune Allnor <all...@tele.ntnu.no> wrote:

On 13 Des, 22:59, Rune Allnor <all...@tele.ntnu.no> wrote:

On 13 Des, 22:06, AGWFacts <AGWFa...@ipcc.org> wrote:
All of the evidence says it is. Why would someone wish to "let go"
of the observed facts?

The *facts* are that the present climate is warmer than
amere couple of decades ago. Any claims that the change
is *man-made* are speculation. As I said in an earlier
post, one can easily find data that show any co-variation,
like between population numbers of humans and storks.

The numbers are facts. Any claim that there is a cause-effect
relation needs to be justified.

The baby-stork example is available on-line, in Amazon's
pre-view of the Box, Hunter & Hunter 2005 edition:

http://www.amazon.com/Statistics-Experimenters-Design-Innovation-Disc...

Don't know how close one gets through that link, but it's
figure 1.3, page 8, under the Look Inside flag.

This is a standard text on data analysis, so it is
rather disturbing that somebody who claim affiliations
with IPCC (the email address you post from) don't
know this.

Maybe just as well you don't post under your real name.

Rune

I didn't know of the stork example. The example I like from astromomy
is in the Northern Hemisphere that total solar eclipses occur more
often on Wednesday than any other day of the week. What could possibly
be the linkage? This fact holds true when looking over 50, 100, 500,
1000 or more years time spans! The astromomer Jean Meus wrote a great
article on this. This is a great example that illustrates statistical
clustering and that correlation does not imply causality.

Clay

 

[Orval Fairbairn]

In article <9krlsgFuv2U1@mid.individual.net>,
Tom P <werotizy@freent.dd> wrote:

On 12/13/2011 07:18 PM, Orval Fairbairn wrote:
In article<9kpfq4FquqU1@mid.individual.net>,
Tom P<werotizy@freent.dd> wrote:

On 12/13/2011 06:10 AM, Orval Fairbairn wrote:
In article<9kngueFj7qU1@mid.individual.net>,
Tom P<werotizy@freent.dd> wrote:

On 12/13/2011 12:13 AM, Tom P wrote:
On 12/12/2011 05:49 PM, Bret Cahill wrote:
Assume the tree ring data is good.

http://joannenova.com.au/2011/12/chinese-2485-year-tree-ring-study-sh
..
.


Bret Cahill

Use of Fourier analysis.

Most would go with the two higher frequencies. The problem is
extrapolating off of the two cycles of the /1300 year frequency.


Bret Cahill





There is another aspect which I cannot understand from the article.
The researchers have presumably found by fourier analysis that there is
some proxy in the tree-ring data that displays the periodic signals
described- all well and good - but how do they determine the
correlation
between the proxy and the instrumental temperature record? By Principal
Component Analysis? If so, why are their results any more reliable than
Briffa's?
Maybe someone with access to the paper can clarify?




Addendum - the paper is accessible, but it refers to yet another paper
for the source of the temperature data for the last 2485 years- based on
tree-ring analysis, lol. Let the paper chases begin!

BTW I'd like to echo Bret's suspicion as well that particularly when it
comes to low frequency signals - meaning cycle time comparable with
sample length - you can prove anything you want with fourier analysis.

.... including "hockey stick" tailoffs. If the data are not smooth, the
FFT will go unstable and show a tailoff at the end of the data stream
that does not represent realistic behavior. A common tailoff is the
"hockey stick" shap, popularized by Mann, et. al.

The end of the hockey stick since 1900 is instrumental data. No Fourier
analysis involved. The disputed part is the pre-instrumental part
derived by statistical analysis of proxies, in particular using the PCA
technique. AFAIK no Fourier analysis there either. Correct me if I'm
wrong.
PCA analysis relies on a time overlap between instrumental and proxy
data and attempts to determine which factors in the proxies correlate
with the instrumental data. It' not at all clear to me why anyone
should think that the extremely small changes in global temperatures
should have a detectable effect on tree-ring growth compared with the
major changes in annual growth due to rainfall or cloud cover, let alone
the implicit assumption that there is a linear relationship, without
which the PCA analysis is meaningless.

You need the Fourier analysis just to sort through the data scatter --
even with "instrumental" data, which, BTW, has not covered the Earth
until recent times.

What you have said so far makes little sense. You said:
quote
.... including "hockey stick" tailoffs. If the data are not smooth, the
FFT will go unstable and show a tailoff at the end of the data stream
that does not represent realistic behavior. A common tailoff is the
"hockey stick" shap, popularized by Mann, et. al.
/quote

But the post industrial data is NOT derived from FT analysis, and
moreover Mann's analysis is not based on FFT

So if you say that FFT "goes unstable", why are you now saying that you
need to use it?

....because you can't get smooth data without it. there is too much data
scatter.

 

[Angelo Campanella]

"Orval Fairbairn" <orfairbairn@earthlink.net> wrote in message
newsrfairbairn-9EFB30.11511514122011@70-3-168-216.pools.spcsdns.net...
.......snip..........

...because you can't get smooth data without it. there is too much data
scatter.

Ergo, the variable (global temperature) is a random variable.... and all
serious students of statistics know that any value is possible for a random
variable including MOMENTARY very low and very high values. The culprit here
is the very thought that such any given HIGH value can remain permanent (hot
days every day). It's absurd to express that in public as the present
Democratic administration ninnies are want to do! Throw them out!

Ange.

 

[fungus]

On Dec 14, 4:37 pm, Clay <c...@claysturner.com> wrote:

 This is a great example that illustrates statistical
clustering and that correlation does not imply causality.

Sure... but relation between CO2 and temperature
isn't a case of somebody going out and looking
for correlation, it follows the scientific method.

* Theory
The Greenhouse Effect was postulated in 1824
by a guy called Joseph Fourier.

* Predicition
"Infrared radiation is absorbed by greenhouse gases"

* Experiment
In 1859 the prediction was proved experimentally
in the laboratory by a guy called John Tyndall.

(And the "ongoing experiment" is of course The Earth...)


Since then the climateologists have been observing
and looking to see if the theory breaks down, it hasn't.

They're also looking for other possible explanations
for their observations. They haven't found anything
that stands up to scrutiny.

Climate change isn't some "Coca-Cola vs. Pepsi"
or "X-Factor" thing where every last idiot's opinion
is valid and we can have phone-in voting to decide
who wins.

The Earth *cannot* warm itself up, the energy has
to come from The Sun (there's nothing else around
here to provide it).

Two indisputable facts:

* The Greenhouse Effect is real, it traps solar energy
and warms The Earth.

* We're pumping huge amounts of greenhouse gases
into the air. They're not being absorbed by trees/algae,
they're measurably changing the composition of the
atmosphere.

 

[Jerry Avins]

On 12/14/2011 11:29 AM, Jerry Avins wrote:

...

The seasonal variation of deaths by drowning the consumption of ice
cream and in the US correlate well. Does this imply causality? Probably
not. Is it to be expected? Almost certainly.

Arrgh!

The seasonal variations of deaths by drowning and the consumption of ice
cream in the US correlate well.

Better yet: In the US, the seasonal variations of death by drowning and
consumption of ice cream are well correlated.

Jerry
--
Engineering is the art of making what you want from things you can get.
ŻŻŻŻŻŻŻŻŻŻŻŻŻŻŻŻŻŻŻŻŻŻŻŻŻŻŻŻŻŻŻŻŻŻŻŻŻŻŻŻŻŻŻŻŻŻŻŻŻŻŻŻŻŻŻŻŻŻŻŻŻŻŻŻŻŻŻŻŻŻŻ

 

[fungus]

On Dec 14, 5:59 pm, "Angelo Campanella" <a.campane...@att.net> wrote:

Ergo, the variable (global temperature) is a random variable....

Fail.

A random variable has no memory.

eg. Roulette is random. On a roulette table
seeing 100 reds in a row doesn't imply the
next spin is more likely to be black. The
probability of coming up black remains
exactly the same as always.

Temperature is not random. Having a hot day
will definitely influence the temperature of the
following day. This is why the highest summer
temperatures are in July-August, not on the
21st of June.

Similarly, having a hot year definitely changes
the temperature of the following year (although
the relationship is a lot more complex).

 

[Rune Allnor]

On 14 Des, 05:49, Bret Cahill <BretCah...@peoplepc.com> wrote:

Assume the tree ring data is good.

http://joannenova.com.au/2011/12/chinese-2485-year-tree-ring-study-sh...

LOL! Good catch. The writer of the propaganda seems to believe the
central-eastern Tibetan Plateau is the entire planet, therefore
all of the scientists on the planet are wrong. Amazing. Funny,
too.

Sure. The same argument works against claims the other
way, that are based on findings on the Greenland glacier
core samples: How would you justify a claim that local
findings are relevant on a global scale?

True but irrelevant here where we humor his claim that his tree data
correlate well with the medieval warm period and other documented
events.
And do contemplate what this means: *several* independent
data sets point in the same direction.

That's irrelevate in this thread as the tree ring data is assumed to
be accurate.

It's very relevant, as few if any data from outside lab
conditions can be taken at face value.

Then feel free to start a thread on it.

The OP here assumes the Chinese had good data.

Sure, the *measurements* might well be good or flawless
or noiseless or whatever. They usually are flawed in
several sense.

The *analysis* is not at all trivial, even with a perfect
data set, when done in isolation. Only when several data
sets have been recorded can one start to see what is
noise or measurement errors in the data themselves,
local variation at a site, and global trends among the
measurements.

 If several
independent measurements point in the same direction,
that tends to be mutually supportive.
Here the problem is claiming a /1300 year frequency is signal when he
doesn't even have two periods of sampling time.  To be somewhat sure
it wasn't noise would require waiting another 1300 years.
Such a claim would be problematic if these data were
all there was.
It's problematic even if the temperature/time graphs are taken as
accurate.  That was the point of the OP.
The point is that you can not evaluate these data in
isolation;

That's not the point of _this_ thread.  See the OP.

The OP did not indicate there was any point,
whatsoever. You ought to know; the first post
in this thread was posted under your name. Either
you need to express your views somewhat more clearly,
or find out who nicked your identity.

in that case they are just numbers. It is
only when they fit the big picture, already sketched
by other measurements, that they start making sense.
There is no way to know if the < 2 cycles of the 1/1300 year frequency
are signal or noise.  You have to wait a few thousand more years to be
sure.
So what?

We don't have several thousand years.

So what?

Unlike climate 'research', *science* works by testing claims,
and expanding on ideas. As I said in a different post, other
people can have a go at these data, or types of data:

- Dendrochronology is well established, tree growth records
are available in large parts of the world.
- Use the same kind of analysis on tree growth records,
and look for similar patterns.
- Tree growth rates are linked to local growth conditions,
the variation for one indivinual tree being dominated
by temperature: Serasonal variations are the reasons
there are growth ring sin trees at all.
- Map growth rates to known local events, like particularly
cold or hot years.

This idea works everywhere, for the time period tree samples
are available. In some parts of the world, tree trunks have
been preserved in marshes and bogs. So if one can find
such a trunk, one can get an idea of local growth
variations during its lifetime. With C14 dating, one can
get an idea about when that life time was.

So you don't need thousands of years, only good, old-
fashioned *scientific* craftmanship.

The FFT or periodogra or whatever is just
a mathematical method you can apply to the available
data.

And when it comes to determining if a frequency is signal or noise,
that method doesn't work for sample times < 2X the period.

As I said, these idnividual data don't prove or
disprove anything. What these people did, is to
do one type of analysis for onr area of the world,
and one period in time. OK, they used the FFT, and
see a strong component at a certain cycle.

It does not prove anything other than that *these*
data show this kind of behaviour. But as I said above,
others can do the same types of analyses for other
parts of the world, and other time periods.

The interesting results will only appear when there
is a large number of such analyses available.

Just have a look at the data available.
As I said early on, the hard part is to let go of
mythlogy, vested interests and presuppositions.

That the deniers' problem.

That's the *scientist's* problem. Thanks for reinforcing
my distinction between climate 'research' and science.

Rune

 

[Jerry Avins]

On 12/14/2011 1:42 PM, Rune Allnor wrote:

...

If you talk about 'instrument temperature' as
'temperature as measured with some thermometer'
then this problem applies everywhere: Climate models
talk about trends on centuries and millenium scales,
while reliable meaurements are ony available for
the past dozen of decades.

This is one of my main arguments against the
climatologists: How can they be so sure, when
they have so little reliable data?

The ratios of oxygen isotopes is taken to be a reliable indication of
global temperature. Types of pollen in sediments indicate the types
plants in a region, thereby indicating local temperatures. There's a
certain assurance when these various indicators all point the same way.

...

Jerry
--
Engineering is the art of making what you want from things you can get.
ŻŻŻŻŻŻŻŻŻŻŻŻŻŻŻŻŻŻŻŻŻŻŻŻŻŻŻŻŻŻŻŻŻŻŻŻŻŻŻŻŻŻŻŻŻŻŻŻŻŻŻŻŻŻŻŻŻŻŻŻŻŻŻŻŻŻŻŻŻŻŻ

 

[Rune Allnor]

On 14 Des, 18:48, fungus <openglMYSO...@artlum.com> wrote:

On Dec 14, 4:37 pm, Clay <c...@claysturner.com> wrote:

 This is a great example that illustrates statistical
clustering and that correlation does not imply causality.

Sure... but relation between CO2 and temperature
isn't a case of somebody going out and looking
for correlation, it follows the scientific method.

You need to read up on the basics. Establishing a
cause-effect relation can not be done by observational
studies. You need to do the experiments.

In order to represent the whole real-worldsystem,
you need to replicate *all* the factors simultaneously,
in the experiment. You can test the effects of CO2,
H20, coulds, methane, and so on *individually* in the
lab, but these results are not valid for the whole
system until you have tested everything at once.

The Achilles heel of climate models is the relation
between water vapour and water clouds:

1) Water vapor absorbs heat and tend to heat up
the atmosphere, when illuminated by the sun.
2) Water clouds reflect inboud solar radiation
back into space, preventing heating from
taking place.

So the key is to understan how, why and where
water vapour switches to droplets, and become
clouds.

Svensmark's criticism against the climate models
is that no climate model is able to predict cloud
formations. That is, no one understand exactly
why clouds for. Sure, there are the condensation
nuklei that cause droplets for form, but where
did those nuklei come from?

No one knows. Svensmark suggests an answer.
Read his 'The Chilling Stars'.

Rune

 

[Bret Cahill]

Assume the tree ring data is good.

http://joannenova.com.au/2011/12/chinese-2485-year-tree-ring-study-sh...

LOL! Good catch. The writer of the propaganda seems to believe the
central-eastern Tibetan Plateau is the entire planet, therefore
all of the scientists on the planet are wrong. Amazing. Funny,
too.

Sure. The same argument works against claims the other
way, that are based on findings on the Greenland glacier
core samples: How would you justify a claim that local
findings are relevant on a global scale?

True but irrelevant here where we humor his claim that his tree data
correlate well with the medieval warm period and other documented
events.
And do contemplate what this means: *several* independent
data sets point in the same direction.

That's irrelevate in this thread as the tree ring data is assumed to
be accurate.

It's very relevant, as few if any data from outside lab
conditions can be taken at face value.

Then feel free to start a thread on it.

The OP here assumes the Chinese had good data.

  If several
independent measurements point in the same direction,
that tends to be mutually supportive.

Here the problem is claiming a /1300 year frequency is signal when he
doesn't even have two periods of sampling time.  To be somewhat sure
it wasn't noise would require waiting another 1300 years.
Such a claim would be problematic if these data were
all there was.

It's problematic even if the temperature/time graphs are taken as
accurate.  That was the point of the OP.

The point is that you can not evaluate these data in
isolation;

That's not the point of _this_ thread.  See the OP.

in that case they are just numbers. It is
only when they fit the big picture, already sketched
by other measurements, that they start making sense.

There is no way to know if the<  2 cycles of the 1/1300 year frequency
are signal or noise.  You have to wait a few thousand more years to be
sure.

So what?

We don't have several thousand years.

The FFT or periodogra or whatever is just
a mathematical method you can apply to the available
data.

And when it comes to determining if a frequency is signal or noise,
that method doesn't work for sample times<  2X the period.

This is what I have said throughout the whole thread:
The data themselves say nothing about what causes
them to look like they do. OK, the present data set
may or may not be subject to a 1300 year cycle.
That's not the point. The point is that the *present*
data, recorded over the past couple of decades, show
variation on a scale that is not at all unique for the
period and location under study.

There is no question climate scientists would like to low pass filter
but there just isn't any time.

Why would they?

Ask the Chinese who did the low pass filtering in _their_ study.

Just have a look at the data available.
As I said early on, the hard part is to let go of
mythlogy, vested interests and presuppositions.

That the deniers' problem.

Bret Cahill

I'd like to try and summarize some of the problems with this piece of
research.

1. The reliability of the tree-ring method.
  Tree-rings do not deliver temperatures. Tree-rings tell us about
annual growth. By using tree-ring data as a proxy for temperatures we
are making assumptions.

2. The proxy data evaluation.

  The paper does not describe the method used but refers to another
paper as source. However, if the same procedure is used as per Briffa,
then most likely the Principal Component Analysis is used.  This in any
case requires an overlap between the proxy data and instrumental
temperature data. It is not clear over what time span reliable
instrumental temperature data has been available in the central-eastern
Tibetan Plateau.  Figure 1 refers to "orange line, calibration series,
464 BC–834 AD; purple line, verification series, 835–1980 AD"
  It's pretty obvious that there is no instrumental data covering the
whole of either of these time spans, so what did they use?

3. Reliability of conclusions from PCA analysis.

  The idea of PCA is that we can set up a model that fits the proxy data
to the instrumental data.  So long as the model is then used for
interpolating, this is unproblematic. However, the model can also be
used to hindcast, meaning estimate past temperatures, as well as
forecast into the future, which means that we are assuming that the
model is valid for time spans outside of the instrumental record.
  Not long ago we had a discussion about PCA in connection with the
"hockey stick" graph. The upshot is that that according to the methods
used to discard or select the principal components it is possible to get
wildly different results. In addition, the Briffa pine-cone analysis
suffered a major debacle when revealed that the resulting forecasts for
years after the calibration timespan were wildly inaccurate. This calls
into question whether PCA analysis of tree-ring data really can deliver
any meaningful results in general, let alone provide a basis for Fourier
analysis.

4. The statistical significance of the conclusions from PCA analysis

  The figure 1 displayed on the website shows standard deviations for
the temperature reconstruction. It is notable that with the exception of
a few points all of the values fall inside the 1 SD limits.
  By comparison, CERN's non-announcement yesterday of the observation of
the Higgs Boson presented data with sigma 1.9 - not enough for a
conclusive announcement.

We now consider the Fourier analysis itself and the presentation of the
results.

5. Exaggerated power spectrum.

Figure 2 shows the power density vs. frequency a-1.  Noticeable is that
the power density is extremely high for values corresponding to 500
years or more. However, this distribution may well be artifact, and is
quite typical for the power spectrum of any noisy data.
  Conversely, the high frequency components are extremely weak.

6. Claimed frequency components.

  As Bret points out, it is meaningless to identify frequency components
comparable to half the timespan of the data itself. Indeed, it is
questionable whether some of the higher frequencies are not simply
harmonics. Figure 3 shows such a picture, so it could be that we are
seeing harmonics of a sawtooth - but could it be we are also seeing
harmonics arising from the "clipping" of 2485 years of data?

By now NASA must have looked at every last thing that has to do with
celestial mechanics, Euler's equations, etc. A 1/1300 year comet or
astroid would be confirmed immediately.

So the question is, what periodic event on earth could take place
every 1300 years?

And if it was solar activity why hasn't 1300 years appeared in some
other record?

It's not impossible but it really seems unlikely this is useful at
all.

7. Presentation of results.

The JoanneNova website chose to present Figure 5 as their favorite
graph, showing as it does a "forecast" of temperatures for the next 120
years(!).
  Interesting is that the 1SD noise present in the original data has
totally vanished. Why no confidence limits?

Maybe the Chinese government finally decided to "go denial" but it's
too late to find any good shills and they don't want to execute their
competent scientists. Maybe they knew it would be exposed but just
wanted to see how long it would take.

In this case the scientist probably wasn't very good.


Bret Cahill

 

[Rune Allnor]

On 14 Des, 15:28, Tom P <werot...@freent.dd> wrote:

On 12/14/2011 05:49 AM, Bret Cahill wrote:





Assume the tree ring data is good.

http://joannenova.com.au/2011/12/chinese-2485-year-tree-ring-study-sh...

LOL! Good catch. The writer of the propaganda seems to believe the
central-eastern Tibetan Plateau is the entire planet, therefore
all of the scientists on the planet are wrong. Amazing. Funny,
too.

Sure. The same argument works against claims the other
way, that are based on findings on the Greenland glacier
core samples: How would you justify a claim that local
findings are relevant on a global scale?

True but irrelevant here where we humor his claim that his tree data
correlate well with the medieval warm period and other documented
events.
And do contemplate what this means: *several* independent
data sets point in the same direction.

That's irrelevate in this thread as the tree ring data is assumed to
be accurate.

It's very relevant, as few if any data from outside lab
conditions can be taken at face value.

Then feel free to start a thread on it.

The OP here assumes the Chinese had good data.

  If several
independent measurements point in the same direction,
that tends to be mutually supportive.

Here the problem is claiming a /1300 year frequency is signal when he
doesn't even have two periods of sampling time.  To be somewhat sure
it wasn't noise would require waiting another 1300 years.
Such a claim would be problematic if these data were
all there was.

It's problematic even if the temperature/time graphs are taken as
accurate.  That was the point of the OP.

The point is that you can not evaluate these data in
isolation;

That's not the point of _this_ thread.  See the OP.

in that case they are just numbers. It is
only when they fit the big picture, already sketched
by other measurements, that they start making sense.

There is no way to know if the<  2 cycles of the 1/1300 year frequency
are signal or noise.  You have to wait a few thousand more years to be
sure.

So what?

We don't have several thousand years.

The FFT or periodogra or whatever is just
a mathematical method you can apply to the available
data.

And when it comes to determining if a frequency is signal or noise,
that method doesn't work for sample times<  2X the period.

This is what I have said throughout the whole thread:
The data themselves say nothing about what causes
them to look like they do. OK, the present data set
may or may not be subject to a 1300 year cycle.
That's not the point. The point is that the *present*
data, recorded over the past couple of decades, show
variation on a scale that is not at all unique for the
period and location under study.

There is no question climate scientists would like to low pass filter
but there just isn't any time.

Why would they?

Ask the Chinese who did the low pass filtering in _their_ study.

Just have a look at the data available.
As I said early on, the hard part is to let go of
mythlogy, vested interests and presuppositions.

That the deniers' problem.

Bret Cahill

I'd like to try and summarize some of the problems with this piece of
research.

1. The reliability of the tree-ring method.
  Tree-rings do not deliver temperatures. Tree-rings tell us about
annual growth. By using tree-ring data as a proxy for temperatures we
are making assumptions.

Sure, but testable assumptions: Subject a set of
sapling plants to different temperatures but otherwise
in similar growth onditions greenhouses for a few years.
Analyze differnces in growth rings. Presumably, the
biologists have already done this, so relations between
groth rates and temperature is assumed known.

No problem at all - that's what's life is like outside
the lab.

2. The proxy data evaluation.

  The paper does not describe the method used but refers to another
paper as source. However, if the same procedure is used as per Briffa,
then most likely the Principal Component Analysis is used.  This in any
case requires an overlap between the proxy data and instrumental
temperature data. It is not clear over what time span reliable
instrumental temperature data has been available in the central-eastern
Tibetan Plateau.

If you talk about 'instrument temperature' as
'temperature as measured with some thermometer'
then this problem applies everywhere: Climate models
talk about trends on centuries and millenium scales,
while reliable meaurements are ony available for
the past dozen of decades.

This is one of my main arguments against the
climatologists: How can they be so sure, when
they have so little reliable data?

 Figure 1 refers to "orange line, calibration series,
464 BC–834 AD; purple line, verification series, 835–1980 AD"
  It's pretty obvious that there is no instrumental data covering the
whole of either of these time spans, so what did they use?

C14 dating? Ancient tree trunks are available everywhere.
Record its growth pattern, do the C14. Average over
hundreds of samples to pin-point ring patterns to
precise intervals.

Apart from that,, written records, particularly in that
part of the world, go a long way back. Some claim
thousands of years. While one would not find statements
like '-10C' in such records, officials will make general
notes as to famines and other problems. So general trends
will be known.

Tax incomes, grain prices, all those economic details
will, when recorded, be very good indocators og general
living conditions. C14 might help to pin-point the
details.

But yes, you point is a valid one: Lack of good data
is a big deal.

3. Reliability of conclusions from PCA analysis.

  The idea of PCA is that we can set up a model that fits the proxy data
to the instrumental data.  So long as the model is then used for
interpolating, this is unproblematic. However, the model can also be
used to hindcast, meaning estimate past temperatures, as well as
forecast into the future, which means that we are assuming that the
model is valid for time spans outside of the instrumental record.

Wrong. The problem is not if the model is valid *outside*
the time span; the problem is if the model is valid *at all*.

But the wise analyst stays away from model-based methods
for precisely this reason.

  Not long ago we had a discussion about PCA in connection with the
"hockey stick" graph.

Mann et al was severely criticesed a few years ago
because no one seemed to be able to reproduce that
graph.

4. The statistical significance of the conclusions from PCA analysis

  The figure 1 displayed on the website shows standard deviations for
the temperature reconstruction. It is notable that with the exception of
a few points all of the values fall inside the 1 SD limits.
By comparison, CERN's non-announcement yesterday of the observation of
the Higgs Boson presented data with sigma 1.9 - not enough for a
conclusive announcement.

You need to be aware of the difference between observational
studies and experimental studies. The objectives, and thus
the tools, differ. This study was an observational study.
Sure, you need to fuse the data from several data sources
into one main trend, but those data can be found from
standard dendrochronology references.

You could argue for or against the exclusion of data points
from known anomalies, like e.g. trees in well-sheltered
areas, that are not subjected to the general climate
variations. Or from trees that are particularly exposed.

But that's part of the scientific method: Explain what
you did, so others can challenge it.

We now consider the Fourier analysis itself and the presentation of the
results.

5. Exaggerated power spectrum.

Figure 2 shows the power density vs. frequency a-1.  Noticeable is that
the power density is extremely high for values corresponding to 500
years or more. However, this distribution may well be artifact, and is
quite typical for the power spectrum of any noisy data.
  Conversely, the high frequency components are extremely weak.

The power spectrum is not the tool for this type of data.
I'd rather use a state space model, random walk + trends,
and maybe use spectrum analysis on the residuals to look
for periodicals.

6. Claimed frequency components.

  As Bret points out, it is meaningless to identify frequency components
comparable to half the timespan of the data itself. Indeed, it is
questionable whether some of the higher frequencies are not simply
harmonics. Figure 3 shows such a picture, so it could be that we are
seeing harmonics of a sawtooth - but could it be we are also seeing
harmonics arising from the "clipping" of 2485 years of data?

Which is why the power spectrum is the wrong tool.

Which in turn brings us to the value of the data set:
It raises questions and inspires ideas of replicating
the same kind of analysis in different parts of the world.
The value of the data is not in the FFT or PCA analysis,
but in the questions they raise.

7. Presentation of results.

The JoanneNova website chose to present Figure 5 as their favorite
graph, showing as it does a "forecast" of temperatures for the next 120
years(!).
  Interesting is that the 1SD noise present in the original data has
totally vanished. Why no confidence limits?

That's the main weakness of the paper: I can't
see, from the material on the website, how one
could make such a prediction, let alone come up
with confidence limits.

Rune

 

[Rune Allnor]

On 14 Des, 16:37, Clay <c...@claysturner.com> wrote:

On Dec 13, 5:16 pm, Rune Allnor <all...@tele.ntnu.no> wrote:





On 13 Des, 22:59, Rune Allnor <all...@tele.ntnu.no> wrote:

On 13 Des, 22:06, AGWFacts <AGWFa...@ipcc.org> wrote:
All of the evidence says it is. Why would someone wish to "let go"
of the observed facts?

The *facts* are that the present climate is warmer than
amere couple of decades ago. Any claims that the change
is *man-made* are speculation. As I said in an earlier
post, one can easily find data that show any co-variation,
like between population numbers of humans and storks.

The numbers are facts. Any claim that there is a cause-effect
relation needs to be justified.

The baby-stork example is available on-line, in Amazon's
pre-view of the Box, Hunter & Hunter 2005 edition:

http://www.amazon.com/Statistics-Experimenters-Design-Innovation-Disc...

Don't know how close one gets through that link, but it's
figure 1.3, page 8, under the Look Inside flag.

This is a standard text on data analysis, so it is
rather disturbing that somebody who claim affiliations
with IPCC (the email address you post from) don't
know this.

Maybe just as well you don't post under your real name.

Rune

I didn't know of the stork example.

I like it, for two reasons:

1) It's available in a standard textbook. Carries some
weight with some, it might not otherwise have had.
2) *Everyone* understand it. One only needs to know
certain extremely basic facts of life, to see the
point. Which means only those who miss those basic
facts of life miss the point.

Rune

 

[Eric Jacobsen]

On Wed, 14 Dec 2011 13:54:37 -0500, Jerry Avins <jya@ieee.org> wrote:

On 12/14/2011 11:29 AM, Jerry Avins wrote:

...

The seasonal variation of deaths by drowning the consumption of ice
cream and in the US correlate well. Does this imply causality? Probably
not. Is it to be expected? Almost certainly.

Arrgh!

The seasonal variations of deaths by drowning and the consumption of ice
cream in the US correlate well.

Better yet: In the US, the seasonal variations of death by drowning and
consumption of ice cream are well correlated.

Jerry
--
Engineering is the art of making what you want from things you can get.
ŻŻŻŻŻŻŻŻŻŻŻŻŻŻŻŻŻŻŻŻŻŻŻŻŻŻŻŻŻŻŻŻŻŻŻŻŻŻŻŻŻŻŻŻŻŻŻŻŻŻŻŻŻŻŻŻŻŻŻŻŻŻŻŻŻŻŻŻŻŻŻ

Knew what you meant, I did.


Eric Jacobsen
Anchor Hill Communications
www.anchorhill.com

 

[Rune Allnor]

On 14 Des, 20:07, Jerry Avins <j...@ieee.org> wrote:

On 12/14/2011 1:42 PM, Rune Allnor wrote:

   ...

If you talk about 'instrument temperature' as
'temperature as measured with some thermometer'
then this problem applies everywhere: Climate models
talk about trends on centuries and millenium scales,
while reliable meaurements are ony available for
the past dozen of decades.

This is one of my main arguments against the
climatologists: How can they be so sure, when
they have so little reliable data?

The ratios of oxygen isotopes is taken to be a reliable indication of
global temperature. Types of pollen in sediments indicate the types
plants in a region, thereby indicating local temperatures. There's a
certain assurance when these various indicators all point the same way.

And which way is that?

These records tend to show, like the article that
set this thread off, that things change, and always
have changed. There is nothing to suggest that
there is anything unusual going on, other than that
people have started doing easurements with the
subsequent onset of mass panic.

Rune

 

[fungus]

On Dec 14, 7:11 pm, Rune Allnor <all...@tele.ntnu.no> wrote:

On 14 Des, 18:48, fungus <openglMYSO...@artlum.com> wrote:

On Dec 14, 4:37 pm, Clay <c...@claysturner.com> wrote:

 This is a great example that illustrates statistical
clustering and that correlation does not imply causality.

Sure... but relation between CO2 and temperature
isn't a case of somebody going out and looking
for correlation, it follows the scientific method.

You need to read up on the basics. Establishing a
cause-effect relation can not be done by observational
studies. You need to do the experiments.

But you're with me that the "stork" thing is
completely stupid...?

That's a start I guess...

 

[fungus]

On Dec 14, 9:13 pm, Rune Allnor <all...@tele.ntnu.no> wrote:

These records tend to show, like the article that
set this thread off, that things change, and always
have changed. There is nothing to suggest that
there is anything unusual going on, other than that
people have started doing easurements with the
subsequent onset of mass panic.

Of course it's not "unusual" that increased
atmospheric CO2 will cause the ice to melt.

The question is whether it's a good idea
for us to be increasing it.

Me? I think it's a very bad idea and I don't see
how any sensible person can think otherwise.

 

[Rune Allnor]

On 14 Des, 21:30, fungus <openglMYSO...@artlum.com> wrote:

On Dec 14, 7:11 pm, Rune Allnor <all...@tele.ntnu.no> wrote:

On 14 Des, 18:48, fungus <openglMYSO...@artlum.com> wrote:

On Dec 14, 4:37 pm, Clay <c...@claysturner.com> wrote:

 This is a great example that illustrates statistical
clustering and that correlation does not imply causality.

Sure... but relation between CO2 and temperature
isn't a case of somebody going out and looking
for correlation, it follows the scientific method.

You need to read up on the basics. Establishing a
cause-effect relation can not be done by observational
studies. You need to do the experiments.

But you're with me that the "stork" thing is
completely stupid...?

Stupid in what sense? The data have been published
in some journal. Did the authors make them up?
The editors of the journal mess up? What do you
mean?

Rune

 

[Jerry Avins]

On 12/14/2011 3:13 PM, Rune Allnor wrote:

On 14 Des, 20:07, Jerry Avins<j...@ieee.org> wrote:
On 12/14/2011 1:42 PM, Rune Allnor wrote:

...

If you talk about 'instrument temperature' as
'temperature as measured with some thermometer'
then this problem applies everywhere: Climate models
talk about trends on centuries and millenium scales,
while reliable meaurements are ony available for
the past dozen of decades.

This is one of my main arguments against the
climatologists: How can they be so sure, when
they have so little reliable data?

The ratios of oxygen isotopes is taken to be a reliable indication of
global temperature. Types of pollen in sediments indicate the types
plants in a region, thereby indicating local temperatures. There's a
certain assurance when these various indicators all point the same way.

And which way is that?

I don't know. When they all point the same way, each imparts credence to
the others.

These records tend to show, like the article that
set this thread off, that things change, and always
have changed. There is nothing to suggest that
there is anything unusual going on, other than that
people have started doing easurements with the
subsequent onset of mass panic.

Winter ice skating was the norm on lakes around here even 40 years ago
and long before that. It hasn't been possible at all for the last 20
years. I don't call that a long-term trend, but records extend it back
in time for another hundred years or so. People who live in the tropics
or the arctic will notice little. Those of us in more temperate areas
see the phenomenon -- whatever it is -- more clearly. For the past few
weeks, many forsythia bushes around here have been showing the
characteristic yellow of their spring foliage. Some of my crocuses,
which often poke up through late-spring snow, are already breaking
ground. Of course, they'll be nipped by the first hard freeze, but that
won't happen for at least a week. I must remember to disconnect and
drain my garden hoses before New Years Day.

Dengue fever has marched steadily northward through Mexico over the last
20 years and has now reached parts of the US. How long before it will be
the scourge in Washington, DC that malaria once was? Tropical fishes are
invading our lakes and rivers. Tropical vegetation is taking over both
water and field. Whatever the cause, warming is not an unmitigated blessing.

Jerry
--
Engineering is the art of making what you want from things you can get.
ŻŻŻŻŻŻŻŻŻŻŻŻŻŻŻŻŻŻŻŻŻŻŻŻŻŻŻŻŻŻŻŻŻŻŻŻŻŻŻŻŻŻŻŻŻŻŻŻŻŻŻŻŻŻŻŻŻŻŻŻŻŻŻŻŻŻŻŻŻŻŻ

 

[Eric Jacobsen]

On Wed, 14 Dec 2011 16:48:41 -0500, Jerry Avins <jya@ieee.org> wrote:

On 12/14/2011 3:13 PM, Rune Allnor wrote:
On 14 Des, 20:07, Jerry Avins<j...@ieee.org> wrote:
On 12/14/2011 1:42 PM, Rune Allnor wrote:

...

If you talk about 'instrument temperature' as
'temperature as measured with some thermometer'
then this problem applies everywhere: Climate models
talk about trends on centuries and millenium scales,
while reliable meaurements are ony available for
the past dozen of decades.

This is one of my main arguments against the
climatologists: How can they be so sure, when
they have so little reliable data?

The ratios of oxygen isotopes is taken to be a reliable indication of
global temperature. Types of pollen in sediments indicate the types
plants in a region, thereby indicating local temperatures. There's a
certain assurance when these various indicators all point the same way.

And which way is that?

I don't know. When they all point the same way, each imparts credence to
the others.

These records tend to show, like the article that
set this thread off, that things change, and always
have changed. There is nothing to suggest that
there is anything unusual going on, other than that
people have started doing easurements with the
subsequent onset of mass panic.

Winter ice skating was the norm on lakes around here even 40 years ago
and long before that. It hasn't been possible at all for the last 20
years. I don't call that a long-term trend, but records extend it back
in time for another hundred years or so. People who live in the tropics
or the arctic will notice little. Those of us in more temperate areas
see the phenomenon -- whatever it is -- more clearly. For the past few
weeks, many forsythia bushes around here have been showing the
characteristic yellow of their spring foliage. Some of my crocuses,
which often poke up through late-spring snow, are already breaking
ground. Of course, they'll be nipped by the first hard freeze, but that
won't happen for at least a week. I must remember to disconnect and
drain my garden hoses before New Years Day.

Dengue fever has marched steadily northward through Mexico over the last
20 years and has now reached parts of the US. How long before it will be
the scourge in Washington, DC that malaria once was? Tropical fishes are
invading our lakes and rivers. Tropical vegetation is taking over both
water and field. Whatever the cause, warming is not an unmitigated blessing.

Jerry
--
Engineering is the art of making what you want from things you can get.
ŻŻŻŻŻŻŻŻŻŻŻŻŻŻŻŻŻŻŻŻŻŻŻŻŻŻŻŻŻŻŻŻŻŻŻŻŻŻŻŻŻŻŻŻŻŻŻŻŻŻŻŻŻŻŻŻŻŻŻŻŻŻŻŻŻŻŻŻŻŻŻ

Nor is it a curse, necessarily, and if it's natural I think it's
beyond folly to try to stop it.

Where I grew up in South Dakota, which is not all that warm
(relatively) in the summer and seriously frigid in the winter, used to
be a tropical swamp according to the fossil records. Retreating
glaciers expose artifacts from people living in the area before the
glaciers formed. Things may well change, and may well change for the
worse, but nature is like that. It seems like it's all happened
before and the earth survived just fine.




Eric Jacobsen
Anchor Hill Communications
www.anchorhill.com

 

[Rune Allnor]

On 14 Des, 21:43, fungus <openglMYSO...@artlum.com> wrote:

On Dec 14, 9:13 pm, Rune Allnor <all...@tele.ntnu.no> wrote:



These records tend to show, like the article that
set this thread off, that things change, and always
have changed. There is nothing to suggest that
there is anything unusual going on, other than that
people have started doing easurements with the
subsequent onset of mass panic.

Of course it's not "unusual" that increased
atmospheric CO2 will cause the ice to melt.

So you attribute all previous changes to changed
levels of CO2?

For the sake of argument, let's run along with that:

1) What natural (i.e. non man-made) sources of CO2
would account for pre-1850 rises in temperature?
2) What natural *sinks* of CO2 would accoubnt for
*falls* in temperature?
3) What would be the reason for discounting the effects
any of the factors listed under 1) and 2), at
present or in the future?

The question is whether it's a good idea
for us to be increasing it.

There are many good arguments against burning
fossile fuels. Global warming is *not* among them.

Rune