Comments on: The "Full" Network http://climateaudit.org/2008/10/06/the-full-network/ by Steve McIntyre Tue, 18 Jun 2013 23:40:26 +0000 hourly 1 http://wordpress.com/ By: Nick Moon http://climateaudit.org/2008/10/06/the-full-network/#comment-165809 Wed, 08 Oct 2008 00:07:07 +0000 http://www.climateaudit.org/?p=4022#comment-165809 Re: #36

Steve: Take a look at the Dongge cave smoothing on the new thread. Is this the sort of effect that you’ve noticed:

Yes the spike at the end really does look very like what you see on an oscilloscope.

Re: #38 – I stand fully corrected.

]]> By: Dave Dardinger http://climateaudit.org/2008/10/06/the-full-network/#comment-165808 Tue, 07 Oct 2008 17:49:50 +0000 http://www.climateaudit.org/?p=4022#comment-165808 Re: Steve McIntyre (#39),

Anyway, please take this to the other thread.

What happened to the other thread? It seems to have disappeared?

[I suppose it will re-appear as soon as I post this, however.

Steve: sorry about that. I was adding some info from UC and inadvertently left it offline. .

]]> By: Steve McIntyre http://climateaudit.org/2008/10/06/the-full-network/#comment-165807 Tue, 07 Oct 2008 15:07:06 +0000 http://www.climateaudit.org/?p=4022#comment-165807 I’m starting a thread on Buterworth filters. Given that frequency fidelity is not an issue in these reconstructions (nobody argues for stable underlying cycles), as others observe, why use this particular method? Anyway, please take this to the other thread.

]]> By: Mark T. http://climateaudit.org/2008/10/06/the-full-network/#comment-165806 Tue, 07 Oct 2008 15:00:55 +0000 http://www.climateaudit.org/?p=4022#comment-165806 Re: Nick Moon (#36),

But I knew someone who did. His standard technique for testing analogue filters was to put a square wave into it, and look at the output on a scope. This allows you to see the impulse repsonse.

Actually that allows you to see the step response (assuming the period of the pulse had sufficient duration), which is different than the impulse response.

Mark

]]> By: Chrisz78 http://climateaudit.org/2008/10/06/the-full-network/#comment-165805 Tue, 07 Oct 2008 09:24:03 +0000 http://www.climateaudit.org/?p=4022#comment-165805 I can say from my experience in a different field (audio engineering) that Butterworth filters are commonly used as low- or bandpass filters when “treating” noisy analogue sound recordings, because of their nice flat response in the frequency domain. In my own audio restoration work I am *not* using them any more however, as I find the VERY noticeable impulse distortion objectionable – a steep BW filter will turn a sharp audio impulse (no matter if caused by a scratched record or by a percussive instrument) into a declining sinusoidal wave of perceptible length, about 5 to 10 times as long as the original impulse, audible as a “ringing” noise like from a dampened plucked string. Definitely NOT what you’d want to use if – like with the temperature curves discussed here – you’d need to preserve the curve SHAPE rather than (a certain part of) its spectral properties! By now, there are linear-phase digital lowpass and bandpass filters available for audio and video use that avoid these defects. Sorry I have no idea how exactly these are implemented mathematically, but IMHO this would be a field where to look for better filtering/smoothing methods.

]]> By: Nick Moon http://climateaudit.org/2008/10/06/the-full-network/#comment-165804 Tue, 07 Oct 2008 09:21:32 +0000 http://www.climateaudit.org/?p=4022#comment-165804 I’m amused by the topic of filter design having cropped up. It’s many years since I did any of this kind of thing and I never really understood the electronics. But I knew someone who did. His standard technique for testing analogue filters was to put a square wave into it, and look at the output on a scope. This allows you to see the impulse repsonse. A Butterworth filter would, if I recall, ring a bit. You’d get a high frequency spike that then decayed very quickly. Bessel filters wouldn’t. They were, in effect, critically damped.

I’m more than a little bit amazed to find this coming up in a discussion on statistics.

But I’m having real trouble imagining what applying such a filter to this kind of data means. If one believed that the climate was a collection of overlapping cycles of different lengths – then this technique could be used to remove the shortest cycles.

But proxies aren’t a continuous signal. They are a series of data points, and the variation between successive data points could be quite large – not disimilar to the change at an end point. If you are getting ringing at an enpoint are you also getting ringing within the dataset. This rather depends on the frequency of sampling (Nyquist?) also on what smoothing might already have been applied.

Steve: Take a look at the Dongge cave smoothing on the new thread. Is this the sort of effect that you’ve noticed:

]]> By: Lars Baath http://climateaudit.org/2008/10/06/the-full-network/#comment-165803 Tue, 07 Oct 2008 07:44:47 +0000 http://www.climateaudit.org/?p=4022#comment-165803 EA and Steve #27:
Maybe that is why the X-ray density sign was swopped. Otherwise the lake sediments would have a negative correlation to other methods!

]]> By: Mark T http://climateaudit.org/2008/10/06/the-full-network/#comment-165802 Tue, 07 Oct 2008 06:27:16 +0000 http://www.climateaudit.org/?p=4022#comment-165802 Causality issues notwithstanding, of course.

Mark

]]> By: UC http://climateaudit.org/2008/10/06/the-full-network/#comment-165801 Tue, 07 Oct 2008 06:18:22 +0000 http://www.climateaudit.org/?p=4022#comment-165801 Re: Steve McIntyre (#21),

Electronics texts are not very helpful on the impact of these filters on noisy stochastic series, which are a different design issue.

If the end-padding method is fixed, RomanM’s impulse response matrix tells all we need ( assuming the processes in question are Gaussian.. )

http://www.climateaudit.org/?p=3504#comment-301058

]]> By: Mark T http://climateaudit.org/2008/10/06/the-full-network/#comment-165800 Tue, 07 Oct 2008 06:12:13 +0000 http://www.climateaudit.org/?p=4022#comment-165800 What is worth investigating is the impact of such an extensive time duration of a 10th order IIR filter.

Mark

]]>