Re: Evolutionist withholds evidence on Haldane's Dilemma

On 28 Apr, 00:58, Tim Tyler <seemy...@xxxxxxxxxxxxxx> wrote:
Josh Hayes wrote:
Walter ReMine <scie...@xxxxxxxx> wrote:
Nunney is withholding key evidence. [...]

I don't think I agree that Nunney is "withholding key evidence" -- in his
paper he says how he constructed his model, and what his results were. You
are free to construct a model according to what he says in the paper, and
if you do, one of two things will happen:

1) You'll get the same results as Nunney did, which will require you to
figure out what the "flaw" is in the model, if you think these results
aren't possible, or

2) You'll get different results, in which case you can write your OWN paper
and publish it.

That sort of stuff is a big waste of everyone's time.

Then you need a third review paper to decide which
researcher is right.

It is far better for everyone concerned if researchers
simply publish their source code in the first place -
preferably with as liberal license as possible.

Then if there's a problem, observers can simply
say: look, a bug on line 96.

Publishing existing programs on the net is usually
inexpensive, so there is not really any good excuse
for not doing it.
--

I definitely agree. Writing programs from first assumptions requires
time and is usually difficult to get the same results, expecially if
numerical and stochastic computations are involved.
In fact, even exculding bugs, there are many things that can cause
problems that alter the results: numerical algorithms have an accuracy
that is tipically depenent on the actual data (the paper might say
"invert matrix A", but there are many algorithms to do and one that
has good accuracy on one class of matrices may give bad results on
another class of matrices).
Stochastical algorithms are always a source of bias: poor pseudo-
random number generators (there are many still around), incorrect
pseudo-random number generator initialization, bias in sampling
algorithms, etc.

I think that everyone publishing some result obtained using a custom
computer program should also publish the source code.


.

Relevant Pages