Merged Discussion of femr's video data analysis
- Thread starter tfk
- Start date
Great graph Tom. I'm gonna link to it when discussing issue with average Truthers when they scream 'freefall' like squawking parrots.
'Freefall = CD! Freefall = CD! Squawk! Inside Job! Larry Silverstein! Squawk!' (parrot shown here with a set of Richard Gage Playblocs for Truthers® - these are gonna be a big hit for Xmas 2006!)
http://upload.wikimedia.org/wikipedia/en/1/12/Alex_the_Parrot.jpg
Yeah, it kinda tells the story.
femr's approach leaves a little to be desired, but it ain't horrible thru the regions of interest. (It is horrible before the building begins to descend.)
The last I heard on this, he was using an absurdly high order (around 50) polynomial.
A far better approach is to fit a series of much lower order polynomials (say, around 3rd to 5th order) over short intervals of the data, matching up magnitudes and derivatives at the junctures between segments. Mathematica or MatLab can implement this easily.
The big advantage of this technique is that all the various segments can be adjusted to fit the local data set quite well, without affecting the whole curve. Much more freedom of choice for the coefficients to match changing frequency events, and the curve can adapt to real discontinuities (such as the start of collapse).
[You can see that femr's curve cannot reflect the very real discontinuity when the building suddenly begins to collapse (i.e., below 12 seconds on this timeline). The acceleration therefore takes off for low values of time (pre-collapse) to numbers like +30 +40 ft/sec^2.]
But that's a relatively small detail.
The interesting point is the shape. Which doesn't match a real free fall at all.
A second significant point is that Chandler - by the very act of choosing to fit a least squares LINEAR regression to the data - FORCED a constant acceleration to emerge from his analysis. Ain't no wonder that he got a constant acceleration. Ain't no accident either. Doesn't matter in the slightest how scattered the data set is. If you fit a linear curve to the velocity, you will ALWAYS get a constant acceleration.
tom
PS. Not surprising in the least that femr has chosen to not respond to my question about the similarity of the curves.
So much for a "seeker of the truth", eh??
NIST said:
Are you saying that the "± 1 inch" in the "13 in. ± 1 in." defines the ±1 inch positional accuracy?
Last edited:
I've looked at a full order range...
Order increases from 1 to 50 in the animation.
Wide-band symmetric differencing has also been performed and the trend is quite similar.
...Ahem...
http://www.internationalskeptics.com/forums/showpost.php?p=6448862&postcount=547
femr2 said:For a period of time (which NIST estimate to be around 2.25s, I suggest a fair bit less) certain points upon the Northerly facade descended between near-to or exceeding g. Outside of that period, acceleration was lower, and not constant for any observable period of time.
A profile of the acceleration curve for the NW corner is...
Different points upon the facade experienced differing acceleration behaviours from that shown above for the NW corner.
Not sure if it warrants a medal, or a stern talking to, but thanks !
Hopefully the noise level will (continue to ?) reduce.
ozeco41
Philosopher
TFK, post 562: "But that's a relatively small detail.
The interesting point is the shape. Which doesn't match a real free fall at all. "
That is you interesting point? That WTC7 does not follow an acceleration step function over a moment you imagine to be "t=0"?
I can tell you really don't know how stupid your comment is. And as for the other readers? Anyone disagree with that?
The interesting point is the shape. Which doesn't match a real free fall at all. "
That is you interesting point? That WTC7 does not follow an acceleration step function over a moment you imagine to be "t=0"?
I can tell you really don't know how stupid your comment is. And as for the other readers? Anyone disagree with that?
You have posted numerous *one liner* posts on this thread recently, none of which have been of any worth, all of which have been less than civil or polite.
In this one you have even chosen to edit my post to make some kind of irrelevant *quip*.
May I remind you of the request from the OP...
tfk said:
Is there some issue you have with my initial response ?
There should not be, unless your intent is some petty attempt to draw out a different response upon which you can pounce with a derision laced *gotcha*. You should be more than capable of answering your own question.
The details in the two NIST quotes I provided you are fairly clear.
Well, as the end result metric is quite significant...
NIST said:
...which refers to the following...NIST said:
The 13 in. ± 1 in. metric is for their *conversion factor*, what we have referred to on this thread as scaling metric, with their uncertainty being +/- 1 inch.
The 14 in. ± 1 in. metric is for their *maximum range of motion* of the west wall edge in the horizontal direction, with their uncertainty being +/- 1 inch.
Are both +/- 1 inch metrics the same thing, clearly not. I imagine you wanted me to say they were.
The important point though Tom...
NIST are stating position change accuracy to +/- 1 inch.
My data for the same movement is clearly of very similar accuracy, and contains significantly less noise.
NIST said:
So Tom, you support the NIST methodology. So, let's see how it relates to my Dan Rather data...
This effectively translates to a +/- 0.1 pixel accuracy.
Via the +/- 4 pixel metric, this relates to scaling metric, which for the vertical axis of the Dan Rather footage...
Vertical Distance - 344ft - 100 pixels
It's over-the-top, but let's say...
100 pixels +/- 4 pixels
3.44 ft +/- 0.14 ft/pixel
NIST seem to be fond of a bit of rounding, so any issue with...
3.5 ft +/- 0.2 ft/pixel ?
Yeah, there is.
I asked very specific questions about what YOU what conclusions that YOU have come to thru YOUR analysis of the WTC7 north wall fall data.
You replied citing by citing NIST's comments.
NIST's comments are not your comments.
NIST's data is not your data.
My issue is your unbroken record of avoiding making the simplest of assertive statements.
And your continuous JAQing off.
Why won't you simply, plainly, directly answer my very simple, very specific questions?
No you didn't. You asked...
http://www.internationalskeptics.com/forums/showpost.php?p=6449263&postcount=563
tfk said:
I though you *understood* the NIST moire method Tom ?
You should not need this explained to you...
NIST basically determine horizontal motion by determining the point along the NW edge that matches their criteria, namely that the pixel intensities either side of the NW edge pixel column are equal. An effect they can infer from what is known about moire.
The NIST metric basically
states that 100 pixel vertical position change of that point equates to 1 pixel of horizontal movement.100 +/- 10 pixel vertical accuracy directly translates to 1 horizontal pixel +/- (10/100 pixels)...
+/- 0.1 pixels
As highlighted, I've looked at the behaviour of the acceleration curve through a full range of degrees...
(1 thru 50)
...and also symmetric differencing upon the raw data.
General trend is similar.
If curve fitting is used at all, I agree that piecewise fitting has significant valid advantages.
The majority is indeed taken up with more and more ridiculous requests from tfk, and my deliberately mild tempered responses. Yes, pretty boring.
I don't really think there is much more to be said on the *call out*. The methods are well honed, and the resultant data of the highest quality, regardless of the hand waving.
It seems any likelyhood of discussion of enhanced noise-treatment processes and such went out the window long ago.
My much improved new Cam#3 data will be available in the near future, utilising the actual video file NIST used, which should improve interest levels.
Given the history of this thread I may decide to either start a new one, or perform analysis at the911forum and simply reference certain observations here, as the unending rubbish from tfk is indeed incredibly boring and ultimately pointless.
I'll let you know when I have the Cam#3 data and observations ready. If you're not interested, no worries.