The 9/11 Forum

David B. Benson wrote:Using the cubic polynomial fitted to my tilt data, I ran my version of the B&V crush-down equation for the first 2.805 seconds of the collapse. As predicted, vaf continues to win out over the BLGB forces, this time by 13.642 AIC units. That is enough to definitely reject the latter forces in favor of vaf.

Could you post corroborating data please ?

femr2 wrote:Could you post corroborating data please ?

Huh? The actual values of AIC are of no interest, just the differences. As always, these are determined by the best fits to the drop data determined from known structure dimensions (also now the antenna mast tilt) and the pixel measurements provided by OneWhiteEye. I actually have no idea what you mean by "corroborating data", sorry.

David B. Benson wrote:

femr2 wrote:Could you post corroborating data please ?

Huh? The actual values of AIC are of no interest, just the differences. As always, these are determined by the best fits to the drop data determined from known structure dimensions (also now the antenna mast tilt) and the pixel measurements provided by OneWhiteEye. I actually have no idea what you mean by "corroborating data", sorry.

How do you perform "I ran my version of the B&V crush-down equation for the first 2.805 seconds of the collapse" ?

I would like to repeat the run.

Enough information for me to replicate the output of "13.642 AIC units" would also be interesting.

Without being able to see exactly how you go about generating the value, or reproducing it, ...

I would have to admit I'm not even sure what AIC means, so i'll include...

.. and what it means

femr2 wrote:How do you perform "I ran my version of the B&V crush-down equation for the first 2.805 seconds of the collapse" ?

On my 'puter, obviously.

I would like to repeat the run.

Aha. This is much more difficult than you might first imagine. Begin by learning some parameter estimation methods.

.. and what it means

Well, I posted this before, but I suppose you didn't see it.
http://en.wikipedia.org/wiki/Akaike_inf ... _criterion

David B. Benson wrote:

femr2 wrote:How do you perform "I ran my version of the B&V crush-down equation for the first 2.805 seconds of the collapse" ?

On my 'puter, obviously.

I would like to repeat the run.

Aha. This is much more difficult than you might first imagine. Begin by learning some parameter estimation methods.

.. and what it means

Well, I posted this before, but I suppose you didn't see it.
http://en.wikipedia.org/wiki/Akaike_inf ... _criterion

I see.

In that case, your staement is entirely worthless to me as you will not corroborate it, explain it in any kind of useful manner, ...

I find the statement "Begin by learning some parameter estimation methods." rather rude, and particularly evasive. You know EXACTLY what I mean by the request, but respond with "On my 'puter". Far from impressed David. You could say, particularly miffed.

I go to effort considerable to provide you with additional resource and detail when you do not grasp what I am saying. Would it not be respectful to respond likewise ?

Please upload every resource with which you perform your calculations, or detail the software required, whatever. If it's bespoke code in an obscure language, not a problem. I have about 30+ under my belt already...if it's a manual process which you follow by hand, no problem, please detail it in full.

It is time for us to all be able to replicate (and so confirm and validate) your stated values, calculations and conclusions.

femr2 wrote:In that case, your staement[sic] is entirely worthless to me as you will not corroborate it, explain it in any kind of useful manner, ...

Had no idea what you would consider to be useful.

I find the statement "Begin by learning some parameter estimation methods." rather rude, and particularly evasive.

Wasn't intended to be either. Do you already know some parameter esimation methods that you trust?

You know EXACTLY what I mean by the request, ...

Obviously not. You didn't 'splain it.

You could say, particularly miffed.

Apologies. This rather pecular form of communication between people who have no good idea of the knowledge base, etc., of the other party does often lead to misunderstandings.

I go to effort considerable to provide you with additional resource and detail when you do not grasp what I am saying.

I do appreciate it.

Would it not be respectful to respond likewise ?

Yes, once I understand something about what you know. I don't wnat to have to remind you how to solve a second order ODE on a computer if you already know. Typically I'll direct you to textbooks if this appears necessary; that's not insulting in the slightest.

Please upload every resource with which you perform your calculations, or detail the software required, whatever.

No. That is not how scientific replication is performed. With some effort you could indeed get SML/NJ running on your machine and then just blindly run it. Which would only show that SML/NJ is indeed machine/os independent, as advertised.

It is time for us to all be able to replicate (and so confirm and validate) your stated values, calculations and conclusions.

Good! The way to do that is to write your own program; that is the computational equivalent of performing a physical experiment in your own lab. We use what is sometimes called computational experiments.

I certainly welcome replication and I'll offer whatever advice is necessary, but it needs to be your work and your program, not mine.

David B. Benson wrote:

Please upload every resource with which you perform your calculations, or detail the software required, whatever.

No. That is not how scientific replication is performed. With some effort you could indeed get SML/NJ running on your machine and then just blindly run it. Which would only show that SML/NJ is indeed machine/os independent, as advertised.

That's how validation is performed in my book. No ? Black box then, the same as NIST ?

It is time for us to all be able to replicate (and so confirm and validate) your stated values, calculations and conclusions.

Good! The way to do that is to write your own program; that is the computational equivalent of performing a physical experiment in your own lab.

I would like to analyse your code David.

We use what is sometimes called computatiional experiments.

A seemingly deliberately rude response David.

I certainly welcome replication and I'll offer whatever advice is necessary, but it needs to be your work and your program, not mine.

Please submit your program for review.

A code-base for Standard ML '97. Simple enough I rekn. No different to me providing the functional spreadsheet model for full inspection.

femr2 wrote:I would like to analyse your code David.

My code is only research quality; it jus' grewed... It would take you less time and verify the result all the better to write your own (and ideally using other than OneWhiteEye's pixel measurements).

We use what is sometimes called computational experiments.

A seemingly deliberately rude response David.

Certainly not intended to be so. I am just using the distinction between a physical experiment and a computational experiment. That terminology is becoming more common.

Please submit your program for review.

No.

Simple enough I rekn.

I fear not, nor is that how scientific replication is performed.

David B. Benson wrote:

femr2 wrote:I would like to analyse your code David.

My code is only research quality; it jus' grewed... It would take you less time and verify the result all the better to write your own (and ideally using other than OneWhiteEye's pixel measurements).

I can fully appreciate the issues associated with ad-hoc code growth. There's no shame. I've performed analysis of some shocking code used in ridiculously mission critical live environments. The result being improvement and progression...

We use what is sometimes called computational experiments.

A seemingly deliberately rude response David.

Certainly not intended to be so. I am just using the distinction between a physical experiment and a computational experiment. That terminology is becoming more common.

My reaction is based upon the distinction between "we" and, what must be presumed "you", aka me.

Please submit your program for review.

No.

Oh.

Simple enough I rekn.

I fear not, nor is that how scientific replication is performed.

Of course I beg to differ. However you have used ML, analysis of the code-base would indeed be quite a simple process. I made it clear that I'm suggesting validation, but you have made your position clear.

Ho hum.

femr2 wrote:My reaction is based upon the distinction between "we" and, what must be presumed "you", aka me.

The we was intended to be all inclusive.

Simple enough I rekn.

I fear not, nor is that how scientific replication is performed.

Of course I beg to differ.

If you are differing with regard to how scientific replication is performed, then I greatly fear you are simply wrong.

I made it clear that I'm suggesting validation, ...

Simply irrelevant to the question of scientific replication of the result, which was what it had seemed to me you were interested in.

Too bad you are not, as it is rather unlikely that anyone else would be interested in doing so.

David B. Benson wrote:

"Simple enough I rekn." I fear not, nor is that how scientific replication is performed.

Of course I beg to differ.

If you are differing with regard to how scientific replication is performed, then I greatly fear you are simply wrong.

Crikey. Here's my un-snipped statement:

A code-base for Standard ML '97. Simple enough I rekn. No different to me providing the functional spreadsheet model for full inspection.

I'm clearly talking about validating the code you have (which you're not going to release.)

I made it clear that I'm suggesting validation, ...

Simply irrelevant to the question of scientific replication of the result, which was what it had seemed to me you were interested in.

Too bad you are not, as it is rather unlikely that anyone else would be interested in doing so.

I think it's clear I've beed asking for resource so I can "check your code and your methods", but hey. I ho-hummed as I'm not fond at all of these, at best, cross purpose discussions.

I may well get around to coding B&V, and if we get differing results, what then ? You say my code is wrong, I say your is ? I'll publish my code, you won't. All a bit irksome. Much less hassle to see where your code is at and try and improve upon it, rather than reinvent the wheel. You've made your position clear.

Ho hum.

Using the cubic polynomial fitted to my tilt data, I ran my version of the B&V crush-down equation for the first 2.805 seconds of the collapse. As predicted, vaf continues to win out over the BLGB forces, this time by 13.642 AIC units. That is enough to definitely reject the latter forces in favor of vaf.

David, many of us may want to understand how you go about the physics. Maybe if we understand the answers to the following short list of questions we can reproduce your results.



Q1: You seem to be trying two models: B&V+BLGB and B&V+vaf, is that correct?

Q2: I assume you are using a particular equation from the BV paper which governs motion and inserting different generalized expressions for force F which contain some parameters. Is this true?

Q3: Are you using eq 12 from BV and trying different forms for the force F on the right side of the equation? If not, which equation?

Q4: Could you please show us or point us to the mathematical form of B&V+BLGB (in the form of a diff eq or two), parameters included, which you are solving?

Q5: Could you please show us or point us to the mathematical form of B&V+vaf (in the form of a diff eq or two), parameters included, which you are solving?

Q6: Is the only difference between BV+BLGB and BV+vaf the form you use for the generalized force F?

Q7: Could you show us the mathematical form F takes for BGLB? For vaf?


You probably posted this before but I never read it on this forum. Answering would help us understand the physical assumptions behind your solutions.

Major_Tom wrote:David, many of us may want to understand how you go about the physics.

Good!

Q1: You seem to be trying two models: B&V+BLGB and B&V+vaf, is that correct?

Yes, after discarding many others.

Q2: I assume you are using a particular equation from the BV paper which governs motion and inserting different generalized expressions for force F which contain some parameters. Is this true?

Yes. In detail I prefer to use equation (22). Indeed, different resistive force functions are used. These contain parameters which the program estimates by best fit to drop data.

Q3: Are you using eq 12 from BV and trying different forms for the force F on the right side of the equation? If not, which equation?

I could use (12), but see the previous answer.

Q4: Could you please show us or point us to the mathematical form of B&V+BLGB (in the form of a diff eq or two), parameters included, which you are solving?

All that is needed in the form of the resistive force function.
F = c + d*S*S
where c and d are parameters to be estimated and S is the speed of the crushing front.

Q5: Could you please show us or point us to the mathematical form of B&V+vaf (in the form of a diff eq or two), parameters included, which you are solving?

F = k*M*S*S
where k is a parameter to be estimated, M is the mass of crushed zone B and S is the speed of the crushing front.

Q6: Is the only difference between BV+BLGB and BV+vaf the form you use for the generalized force F?

Yes.

Q7: Could you show us the mathematical form F takes for BGLB? For vaf?

See above.

Yes. In detail I prefer to use equation (22). Indeed, different resistive force functions are used. These contain parameters which the program estimates by best fit to drop data.

Fair enough. Do you see eq 12 and eq 22 to be the same equation governing motion, just written is different forms?

F = c + d*S*S
where c and d are parameters to be estimated and S is the speed of the crushing front.

Do the stars in the equation have any special meaning (like complex conjugate or so)? Any use of imaginary numbers? (It's just the symbols that I associate with complex numbers. Just want to be clear.)

If not, it's just F=c+ds^2, no?

F = k*M*S*S
where k is a parameter to be estimated, M is the mass of crushed zone B and S is the speed of the crushing front.

.

Again, involving complex numbers? If not, it's just F=kMs^2, no?


I know some of these questions are redundant, but just to be clear:

Is it true that what you mean by BV+BLGB is just BV eq 22 with F= constant plus a term which is proportional with the square of the speed?

Is it true BV+vaf is just BV eq 22 with F proportional to both the instantaneous mass of zone b (=M) and the square of the speed?

Major_Tom wrote:Do you see eq 12 and eq 22 to be the same equation governing motion, just written is different forms?

Same but for the units used.

it's just F=c+ds^2

Yes.

it's just F=kMs^2

Yes.

Is it true that what you mean by BV+BLGB is just BV eq 22 with F= constant plus a term which is proportional with the square of the speed?

Yes.

Is it true BV+vaf is just BV eq 22 with F proportional to both the instantaneous mass of zone b (=M) and the square of the speed?

Yes.