DNAunion

DNAunion: So? It is there for flexibility. If someone wants to change the model to use replacement (instead of discarding objects after they’ve already been selected), he/she simply changes a single value and the whole program is “reconfigured” – it works under the new set of parameters without requiring any new code to be written, or any existing code to be commented out.

DNAunion: I will point out that it is considered, by a great many programmers, to be a bad idea to sacrifice clarity for compactness.

When a tile is chosen, there are two possibilities: either it has been chosen before, or it hasn’t, and both possibilities have to be taken into account. If the statement setting nFoundOne to 0 were omitted, it could cause temporary confusion on someone encountering the code for the first time. “What? The programmer left out the code that handles the case where the tile was already chosen”. Putting it in makes the logic clearer.

As far as compactness:

1) It’s a only 14 characters and took me only about 2 seconds to type (and as the programmer, I get to decide whether to sacrifice clarity to save 2 seconds worth of typing)

2) The code does not become “fatware” simply because I added something like 14 bytes to it

3) The time it takes an average computer to execute that one command is negligible; even running 1,000,000 iterations it would probably add only a small fraction of a second to the total execution time.

So again, it is a tradeoff, and it is up to the programmer to decide how to handle it. That you would have done it a differently hardly makes the way I did it wrong, flawed, or faulty.

DNAunion

DNAunion: Dodging the issue, huh Principia.

Come on you great programmer you. Show us how you would have a computer "make a decision", such as whether or not a tile had been chosen before, and act on that "decision" appropriately, without using any kind of conditional branching.

Gee, I bet Principia ignores or dodges the issue again! LOL!!!

Principia

No, the flaw in your code is not the same as the inefficiency in your code. I have demonstrated both a flaw and inefficiency. Strange, you know, considering how you have agreed with what I said both times, despite my being not a programmer and all.

Shall we continue this programming tutorial in an evolution/creation forum? Because I've already told you why I am not jumping through more hoops for you (look 4 posts above this one).

DNAunion

DNAunion: Wrong.

Quite clearly, you don't understand how the comparison in table 8 was done. It takes the prokaryotic proteins - you know, the things that aren't dipeptides - and sees how many matches the SIPF reaction produced. So you can't completely ignore the prokaryotic proteins, no matter how much you want to.

I am discussing the archaebacteria amino acid pairings and how both A-B and B-A should be present, even if they are the same pair (such as Ala-Ala), because two separate facts are being presented: in the archaebacterial proteins, which amino acids frequently PRECEDE Ala and which amino acids frequently FOLLOW Ala. As I have clearly shown, these are two different things, even if the same amino acid appears in both cases.

Principia

Good for you, but that's not the problem in your model that I am pointing out. As I predicted, here comes the 6th attempt. You are counting coincidences for a homodimer in both the A-B and B-A categories, when they are in fact the same reaction. I don't care about the heterodimers.

And now for the 7th attempt...

Principia

LOL, counting characters to determine computational complexity. That's good. Why I guess I can add this somewhere to your code: and that wouldn't make it "fatware" at all. After all it's only 14 characters too.
So? Whatever happened to flexibility and reusability and empowering users? In other words why do you care about only 1,000,000 iterations? After all, I don't see any evidence at all that 1,000,000 iterations suffice to produce a suitably accurate results. Suppose one of those "empowered" users demands 1e18 iterations? What then?

PS: Yes, as the programmer (or shall we say "designer" ) you get to do anything you want. But, in the end, you are the not only person judging your work.

DNAunion

DNAunion: Stop looking only at dimers and start looking at the archaebaterial proteins!

I've told you – like 10 times now - that the ARCHAEBACTERIAL Ala-Ala has to be included in both A-B and B-A sides of the table, and that is correct.

Now, if one is going to compare the archaebacterial pairs with the same pairs from the SIPF reaction to see how matches occur, then that person has to compare archaebacteria A-B Ala-Ala to SIPF A-B Ala-Ala, and archaebacterial B-A Ala-Ala to SIPF B-A Ala-Ala. Since the dipeptide Ala-Ala is the same for A-B and B-A – which is NOT the case for the archaebacterial proteins – then it seems to me that some of the main possibilities are to either:

1) Just live it and use the single Ala-Ala value for both A-B and B-A, as Rode did in his paper.

2) Split the Ala-Ala value in half to take into consideration that it represents both A-B and B-A (which I suggested, even though it makes no difference for Ala-Ala)

3) Chuck the whole thing.

Since my point was to show the error in Rode’s calculation, which I did (as Principia pointed out), 3 is not an option. That leaves 1 and 2.

I suggested 2 but it didn’t make a difference for Ala-Ala. As far as any others…who cares! My point was to show that Rode’s calculation was way off – and I did so (as Principia pointed out).

So what's the big deal about my following in Rode’s footsteps and using 1? Nothing really. With or without this “Ala-Ala” problem, I showed that Rode’s calculation is multiple orders of magnitude off. I killed his calculation, and I needed to do so only once, in only one way. And that I did (as Principia pointed out).

pz

I would rather not, either. However, in the brief time since I posted that warning you guys have ping-ponged back and forth 9 more times. This is the last time; instead of rushing to get one more dig in against the other guy, stop and gather your ideas into a post with more substance and less sniping.

I'm going to close the thread for 24 hours. You both can use that time to organize your arguments a little better.

pz

OK, this thread is open for business again.

Play nice, everyone.

Principia

Thanks to pz for reopening this thread. It seems to me that the readership has dropped to nothing, so I might as well write up my closing arguments, and let DNAunion have the last word. In his last post yesterday, DNAunion complained:
And this premise is patently false. Look, there are many ways of saying this, but it seems like I have but tried them all. Perhaps a visual demonstration is in order. Suppose I give DNAunion the following sequence: D-E-Ala-Ala-R-Q in an archaebacterium, how does one tell whether or not the Ala-Ala is an A-B linkage or a B-A linkage a posteriori? Hell, why ask when the fact of the matter is that Rode does not distinguish between these for either SIPF dipeptides or the archaebacterial proteins. Logically speaking, it is the only responsible thing for Rode to do, since they provide no means of determining whether an Ala-Ala is an A-B linkage or a B-A linkage. But, in the end what Rode reports in his paper is all that matters for this discussion. One last time: the prevalence of homodipeptides is not distinguished by Rode for either SIPF or archaebacterium or prokaryotes. One only has to look at the data I posted on pg. 2 of this thread, to see that there is one diagonal for homopeptides in each of the SIPF, archae, and prok tables.

It is beginning to be clear that DNAunion is making most of this stuff up. Once again, I urge him simply to read the paper. Perhaps we should have DNAunion answer a pop quiz on the paper, with these trivial questions: How many total residues were used to tabulate the archae/prok data? How many different proteins did these residues represent?

DNAunion "killed" Rode's calculation, and I killed DNAunion's model. As I have been pointing out repeatedly since my first post on this thread, it doesn't really matter whether or not DNAunion comes up with a different number. First of all, as I have said on numerous occasions, whatever it is that DNAunion killed seems to keep resurrecting itself: Hmm... not a very decisive kill at all, when DNAunion can't even say that 1e-7 invalidated Rode's conclusions. Secondly, as I mentioned earlier, touching up a flawed model doesn't mean anything at all. Rode's data, while compelling on its on merits, did not deserve the shabby probabilistic model that he came up with. Two points on this matter: a) Rank-ordering throws away valuable information about the actual magnitude of the data. Rode thinks that simple correlation studies are not possible, but that does not preclude other statistical studies; b) Double counting the same data twice is a no-no. And as I pointed out above, not even DNAunion's touch-up fixed this problem.

So, what's the real agenda here? Well, from DNAunion's posts, it is quite clear that what he wants is for somebody to stroke his ego -- all for two pieces of code, one of which as we later find out was never properly debugged, and the other which was flawed. Or as he put it begrudgingly: So, DNAunion might have expected me to lower my standards for him and to lavish great praise on him for such shabby work, but of course he'd be wrong. In point of fact, because I was unwilling to give what he considered his dues, he taunted me to demonstrate that his code was useless and to find flaws in his code. I generously gave my time to each. First, I showed that in fact, a Monte Carlo analysis was a complete waste of time. I gave two demonstrations of one-line calculations that produced the exact results, yet took DNAunion a million iterations of bloated code to figure out, and even then only to some unknown degree of accuracy because of a flawed code. Two demonstrations later, he is still asking me to do all of the remaining calculations, even though I repeatedly told him that I don't believe in Rode's model as presented. He makes these requests even though he didn't bother to verify my calculations -- clearly he doesn't care. Well, DNAunion may want to learn to think for himself, rather than taunting people to do his work for him. Anyway, after demonstrating that a Monte Carlo analysis was irrelevant, I showed him clearly where sections of his code defeated his proud claims that it was 'resuable,' 'flexible,' and 'empowering users.' I won't rehash these arguments, but clearly, DNAunion's programming style leaves much to be desired. Clearly defeated by my critique, DNAunion put up a weak defense: Well, we all know how that story turned out. I'll let DNAunion have the last word in defending his own code. The matter is only peripheral to the discussion on Rode's paper.

So, at this point, what do we make of the discussion? Well, let's go back to something DNAunion wrote: This is a good place to remind ourselves how science is typically done. I think it is quite clear by now (if not to DNAunion, then to the readers) that Rode's probabilistic model is incomplete and uninformative. So from these 3 alternatives that DNAunion followed, what is the best route to take? Choice 1) still leads us to a broken model. There is no way around this consequence. And I am not talking about just the double-counting problem. I am talking about how the rank-ordering throws away information for the sake of making it look like there are coincidences. Choice 2) is completely nonsensical, since Rode was unable to distinguish between Ala-Ala values that were either A-B or B-A linked. That leaves us, of course, with 3). And that is what a good scientist ought to do when confronted with an unsalvageable model: chuck it. The data is still there. We just find another way of explaining them. What DNAunion is suggesting when he says: is that he affirms all of the creationist scientists who insist on following in the footsteps of Biblical literalists, who to this day can only make incremental progress ... using the wrong model.

Science is in absolutely no rush to solve all of the mysteries of life. Certainly, it doesn't look to flawed computer programs to ascertain the truth. I'll end by going back to what Art briefly mentioned back in page one that there was in fact no competing model for which to compare results. I think that is a good place to start.