Comments on: EMD Estimates of Natural Variation

By: davids99us

davids99us — Fri, 13 Nov 2009 07:03:09 +0000

Sorry you have to go, I would like for someone to prove that it is wrong to conclude that EMD shows natural variation exceeds the trend over the 1975-2000 period – which you have not done.

The assumption that the residue is a all due to trend and not part of another cycle is designed to help AGW out maximally. As I said before, the residue could in fact be part of a longer cycle, but by assuming the trend is all due to AGW, it shows the maximum contribution that AGW could be making. And according to EMD, it is less than the contribution due to natural variation — therefore IPCC lies.

By: isnowball

isnowball — Fri, 13 Nov 2009 06:47:24 +0000

I have other fish to fry . Check your assumptions.

By: isnowball

isnowball — Fri, 13 Nov 2009 06:46:57 +0000

Since the signal is non-stationary, you only have a small piece of a non-linear signal of the signal to work with – hence you're setting yourself up to be fooled.

By: davids99us

davids99us — Fri, 13 Nov 2009 04:12:27 +0000

In you rebuttal you might like to respond to this paper in
PNAS.(http://rcada.ncu.edu.tw/reference008.pdf) with the identical
analysis, except for stating that the relative proportion of NV exceed
the residue trend over the 19975-2000 period.

"As discussed above, regression, moving mean, and filtering all
are problematic in dealing with nonlinear nonstationary data.
With these considerations, only the recently developed EMD
method (1, 5–8) fits the requirements."

By: davids99us

davids99us — Fri, 13 Nov 2009 03:36:28 +0000

So you say, but have yet to say why.

By: isnowball

isnowball — Fri, 13 Nov 2009 03:21:38 +0000

It's clear you don't understand the implications of a non-stationary signal.

By: davids99us

davids99us — Fri, 13 Nov 2009 02:58:46 +0000

The purpose of the method is to extract out the stationary (ie
cyclical) components of different period, leaving what EMD calls the
residual. Now the residual could be a trend or could be a long period
cycle. The debatable issues are over the nature of the residual, and
the end point constraints. As you can see from the graphs, the
residual in the period form 1975 to 2000 is almost linear anyway, even
though it ultimately does appear cyclical, so makes little difference
to the comparison of the amound of natural variation (ie stationary)
vs trend over the period.

By: isnowball

isnowball — Fri, 13 Nov 2009 02:34:45 +0000

The signal is non-stationary and it's non-linear.

When you model a non-linear signal with a linear model, then model has no predictive power, i.e., it's lagging indicator.

And when you only look at a signal which is not stationary, i.e., when you're only looking at part of the signal, then you're setting yourself up to be fooled.

By: davids99us

davids99us — Thu, 12 Nov 2009 21:08:07 +0000

But that is exactly why it is good for natural variation. I don't
follow you objection.

By: cinaedsimson

cinaedsimson — Thu, 12 Nov 2009 20:40:11 +0000

Evidently, you don't seem to realize that EMD assumes the signal is stationary and nonlinear.