Measurement Systems 101 - A P4T/CIT Tutorial or NoC Verified

[BCR]

One aspect of modeling the final seconds of AAL77's path based on the available radar and flight data recorder data seems to escape the fans of P4T and CIT on an on-going basis. So, I thought I would do a very basic primer on what is meant by a measurement system and the range of error associated with it.

For this primer, the measurement system is actually two systems interacting with one another. The first is Google Earth, and the second is the human being using the Google Earth ruler tool.

In the image above, I used the GE ruler to measure the distance from a corner of the Citgo station canopy to a corner of the Pentagon. Obviously, the distance is 1420.73 feet, right? Wrong. If any one of you take the same measurement, the odds are that you will come up with something slightly different. So, which is it? Is it the value I got, or the value you come up with? The answer is neither, and both.

If I start all over and measure the distance again, I will come up with a different result. So for this primer, I took the same measurement 50 times.

If I take those 50 measurements and create a histogram using the Excel data analysis tool, then I get the results above. There is a 9.87 foot difference between my largest measurement and my smallest. The histogram definitely exhibits the bell-shaped normal distribution curve, but with a slight skew to the right, with a mean of 1421.47 feet and a standard deviation of 2.28 feet.

What the histogram tells me is that I can be 95% sure (2 sigma) that the actual distance (assuming that GE is accurate) is 1421.47 +/- 4.56 feet. If I look at the 50 measurements I made, I can expect to find ~5% of them to fall outside the predicted range. Indeed, the 95% range is 1416.91 - 1426.03 feet, and the actual measured range is 1416.61 - 1426.48. Very darn close!

If I want to improve my prediction so that I can be more certain of my estimate, then the 3 sigma range (99.7%) can be used, or 1421.47 +/- 6.84 feet, or 1414.63 - 1428.31 feet. Comparing this predicted range to the actual measurement set finds that ALL of the measurements fall within this range.

Different people (measurement systems) will have different mean values and standard deviations associated with their results. By doing a correlation of the results from different people, a much better idea of the actual distance can be gained, but there will yet remain a range of uncertainty which must be estimated and given.

Now, the error made by P4T and CIT is that in estimating the flight path, they have used almost exclusively eyewitnesses on the north side of the flight path mean as determined by radar/fdr. So I did the same thing for my measurements, I repeated my calculations using ONLY the measurements made greater than the 50 measurement mean (1421.47). The result was a mean of 1423.25, which is greater than the complete set mean (or north of the 'path').

The P4T/CIT boys have excluded for whatever reason ALL eyewitnesses to the lower (South) side of the data mean, so the inevitable result is that their mean is greater (North) than the data mean. The paths draw by those witnesses diverge north of the mean as expected with convergence at the mean at the end of path. If they used the opposite approach (as they did with eyewitnesses west of the Sheraton), the opposite result can be expected and is indeed the case in the CIT flight path reconstructions in that area.

So, short and sweet, CIT is getting exactly (NoC) the results that would be expected from the sampling method used. I hope this has been helpful to the P4T/CIT boys and we can all now expect a new video from them using ALL of the eyewitness accounts to create a flight path reconstruction.

 

[dropzone]

Sets? You're talking sets when the last time I dealt with them, before program-design classes in college, was in Sputnik-era, Red-Scare, 4th-grade math classes?

 

[BCR]

I'm just trying to be helpful

 

[Baylor]

Aldo knows as much about confidence intervals as he does about diet and exercise.

 

[tfk]

I'm just trying to be helpful

That is helpful. And meaningful.

Now, for your next lesson, could you do "precision of measurement & significant digits"?
And you might want to toss in how "screen resolution" plays into it.



Tom

 

[tfk]

Sets? You're talking sets when the last time I dealt with them, before program-design classes in college, was in Sputnik-era, Red-Scare, 4th-grade math classes?

The sets were black & white, had funny metal things on top (frequently replaced with bent coat hangers), and got a grand total of 3 channels.

The "remote" was the youngest sibling in the room over the age of 4.

But the cool thing was, once you bought the set, you didn't have to pay some company $70/month for it to work.

 

[sheeplesnshills]

So, short and sweet, CIT is getting exactly (NoC) the results that would be expected from the sampling method used. I hope this has been helpful to the P4T/CIT boys and we can all now expect a new video from them using ALL of the eyewitness accounts to create a flight path reconstruction.

Not convinced your analogy is quite right but you are on the right track. CIT did interview several SoC witnesses but discarded their testimony as it didn't fit their fantasy. They also distort the evidence of several others, principally Paik and Morin, to make what they say fit an NoC when the first was in no real position to say where the plane went and the latter is quite clearly SoC once you know where he was and what he says and could have seen.

The whole CIT idea fall apart when it depends on the plane not having hit the Pentagon. Many saw it hit, no credible witness saw a flyover and many other witnesses in excellent positions to see one did not.

The reality is that CIT and P4T are just in the fleece the nutcases business

 

[BCR]

Not convinced your analogy is quite right but you are on the right track.

You don't have to be convinced. The 'laws' for sampling and distribution apply whether you are talking about positions 'measured' by eyewitnesses, radar, INS, or a scanning electron microscope.

 

[sheeplesnshills]

You don't have to be convinced. The 'laws' for sampling and distribution apply whether you are talking about positions 'measured' by eyewitnesses, radar, INS, or a scanning electron microscope.

Yes but honest people don't remove any data that falls outside the wished for result or pretend points are not where they were to start with! One cannot really apply your analysis to falsified data.

 

[BCR]

Yes but honest people don't remove any data that falls outside the wished for result or pretend points are not where they were to start with! One cannot really apply your analysis to falsified data.

Ah, but that is the point. At the end I excluded data on one side of the mean to get a skewed result. That is exactly what CIT did.