Back from lunch.

I think we are mostly agreeing, so I have to blame myself for being unclear.

I agree that model should (almost) always start with experimental results and always end with experimental results. But, we have access to very powerful computers, and it seems foolish not to use them to get from the initial experiment to the final experiment.

When I finish my thesis, I'm going to compare my results to experimental results. If they don't match up, I know I've done something terribly wrong. But, if they accurately model current experimental results, my hope is that my model could be used to model *new* things, so we know what to expect before running the experiments.

@red

I think it depends on the field. for instance, wasn't QM developed almost purely as theory and then confirmed by experiment. Conversely, the Transistor was developed experimentally and then they theory came much later. It all depends.

I'm not sure what you mean by Monte Carlo vs. calculations. MC is used because, in a lot of cases, analytic solutions to equations are impossible.

I suppose we are agreeing then.

@fulhamish
I'm not sure about the answer to your question, but I believe there are various problems when modeling kinetics compared to thermodynamics. Thermodynamics can for a large part be modeled in terms of equilibria ("stable states") wheras kinetics deals with transitions from one system to the next ("dynamic states"). I believe most of the difficulties that are encountered when modeling kinetics that are not experienced when modeling thermo can somehow be boiled down to this.
@abgemacht I agree completely; letting a computer do a "test" (let's call this the operation before the experiment) in a fraction of a second to set up an experiment that can take ages to perform is a very wise course of action and one not available earlier, possibly even putting some "ground rules" of science into question.
Second point: Monte Carlo simulations are a series of "modeled" experiment, where you somehow put in a random number, which mimicks the randomness of any experiment. The wiki on the topic gives a nice explanation and some nice examples.
My question would be whether a Monte Carlo simulation can conceptually be put in between <model> and <experiment> for this reason, or whether it is an independent way of studying science.
My personal answer is that although conceptually you can put them in between the two, it's always a bad idea to use them anyway :)

"My question would be whether a Monte Carlo simulation can conceptually be put in between <model> and <experiment> for this reason, or whether it is an independent way of studying science.
My personal answer is that although conceptually you can put them in between the two, it's always a bad idea to use them anyway :)"

Not really sure what you mean by this at all. Especially the bad idea part.

I'm with Yonni. I just don't understand what you're saying in regards to MC.

MC is a modeling method. It should be considered part of the <model> section. What exactly don't you like about MC?

Yonni,
I thought it would be interesting to have some starting point on a debate on Monte Carlo simulations, as the topic popped up several times and it definitely qualifies as a "modern numerical method to science". The randomness involved is supposed to mimick the experimental spread you always find when doing an experiment and the average value obtained from your simulations would be a way to determine "the experimental value".
My question is whether that means that the MC method is a compromise between experiment and model, and if so, if it is a good compromise.
My opinion is that it is in fact a compromise, but not exactly the best one: I'd describe it as "the best way to not have to do science".
I felt brain dead after first trying to perform MC calculations and only when I started to think about the problem I was dealing with at a more fundamental (read: model) level I obtained something of an explanation.
But maybe you have a different experience with it, in which case I'd love to learn.
Redhouse

@ab: I only take my personal experience as a reference point, but in my experience the MC method did not deserve the <model> status: too much honor, basically I was trying to *avoid* having to model, rather than model. But this is a discussion I'd be interested in having and judging by what you said so far you disagree with me, making the discussion the more exciting :)

I'll admit I haven't used MC before, but people in my lab have and it seems to work nicely. I'm going to look into it some more.

But, from my current understanding, I wouldn't think of MC as substantially different than any other numerical technique. There are many numerical methods that use "randomness" as inputs to help get the answer. Furthermore, there are many systems, such as noise and many parts of QM, that genuinely appear to be random; wouldn't it make sense to use a model with "randomness"?

Well in theory yes of course. But in practice (my personal practice) the MC simulation was not so much a simplification of either the experiment or an extension of the model, but a banalisation of both. I want to know if there's people out there who gained real actual *insight* with the method. Because I only got insight when I tackled the same problem using a mathematical formalism.

"Furthermore, there are many systems, such as noise and many parts of QM, that genuinely appear to be random; wouldn't it make sense to use a model with "randomness"?"

This.

The applications I was thinking of involve describable random behavior. The randomness is not modeling uncertainties in an experiment but, rather, the actual behaviour of particles.

@red

Did your experiment involve a lot of randomness? By randomness, I don't mean noise, but randomness inherent to what you're studying. For instance, the location of an electron is determined probabalistically. If not, then I can understand how MC may not give you insight. If so, then I'm not sure what the problem was, as you're modeling the system as true to the real world as possible.

Interesting, for myself I hadn't formulated the difference so clearly. +1 Yonni.

Yeah, I guess you guys are right. Randomness in experiments can be both a result as well as a cause. +1 to ab as well.

Well, I'm glad we can at least agree on that : )

By the way in my head I'm going back to the simulations we did and now that I think about it we definitely used the randomness as input, but we sought to calculate the spread on the output. That didn't work. Thinking about it this way clears it up, but it's a pity when I worked on the problem I worked for a company and I was under a secrecy agreement.

If we've settled this, I would like to go back to one of your original points:

What were you referring to when you talked about the double standards with new nations taking an interest in science?

That would be more of a "state of education" topic. I don't think it belongs in this thread anymore. I would personally be more interested in going back to the original point of the discussion that discusses how science has evolved (MC simulations are an interesting example of that, especially if I think about it the way you guys proposed) over the last years and how it changed the way we look at the world.

I think there are 2 noticeable ways science have changed and they are both thanks to computers:

1) Calculation--We can perform calculations/models faster than anyone 100 or even 50 years ago could possibly imagine.

2) Communication--We can instantly communicate with scientists around the world. Making things more accessible with online journals and repositories such as arXiv helps connect science in a way that was not possible even 25 years ago.

I think in the case of 2 - but after this I have to put in some numbers in my computer or I will not be happy tomorrow :) - the recent "discovery" (?) of an elevated light speed is not a purely (although definitely mostly!) positive development. That should have been dealt with much less openly because it potentially discredits physics as a field: people with no background in science don't understand what's happening.

The first one is solely positive. Does it change our perspective on things, I wonder? Give us the idea that somehow we are in control of the world that we now can calculate a lot from (the climate debate would kick in here I guess)

*Discuss* :)

Oops: what I meant to say was that 2) is not a purely positive development given the consequences: the light speed experience.

The public at large have always and will always misunderstand science. I don't see any way around that.

The OPERA/CERN experiment *was* kept to themselves for months and they were unable to find an error in their experiment. Personally, I doubt very much that neutrinos go FTL, but we shall see. I'm talking about communication between scientists, which is always a plus, no? Communicating science to the public is a separate issue entirely.

Yeah I was already guessing that the problem was more with the media that reported it than the scientist who asked help for the interpretation. But still, you know that once something like that comes out, the media are going to explain your .00000001% higher value as "undermining the theory of relativity", even though that's a nonsense interpretation....

The media will misinterprete anything (Y2K, swine flu, bird flu, etc.)

But, I am very interested in that 2nd bullet point you made. If you think it'd be off-topic, we can make a new thread, but I really want to know what you meant.

No it's no need to make another thread, nobody's listening anymore anyways.
It may be a local thing, and it is not related to science as much as it is to education, but since you insist I'll clarify.
In Holland we are "lacking scientists", in other words; we need more brains that supply the technology to keep our Shells, Unilevers, Philips' and our Universities competitive, after all we're proud to have such huge companies and respected Universities.
What we always did was: let all kids take a national test at age 10 (we still do that) divide them in different types of schools, those who learn faster get a more challenging program than those who don't learn as fast, and if you manage the highest level you can go to University. Now, we're filling up these Uni places much with people from developing countries. While they in their countries passed their own "national tests", they often appear to disappoint (I didn't often manage to have a sensible conversation with them in English). The point was that for some Dutch (very bright) students it was hard to get Ph.D. positions that were awarded to hard-working Asian kids, that did what the professor wanted, but didn't have the "critical autonomy" of the kids trained within our system.
So I am not sure whether the influx of people from different cultures has such a benificial effect on education, unless they are trained in , for example, Holland, from a very young age.
So.. let the criticism that i'm racist begin ;)

I study Physics, so I want to say something about this. The compatibility betwenn the hypotesis of neutrino slower than light and the data observed is over 6 sigma, and this it means that there are 2 chance in one billion that neutrinos observed are slower than light..., but even if there are no mistakes in measurement and neutrinos are FTL, relativity is not undermined, because it was confirmed in many other ways that it's impossible it is wrong, it will be simply an approximation of a more precise theory, not valid in general but under particular conditions, such as Newtonian theory. And if OPERA esperiment is wrong, we have however earned the poor figure of italian ministry of education, with the "tunnel that leads from Ginevra to Gran Sasso".

Ahh... Ginevra may be .000000000001 cm closer to Gran Sasso then? :)

@redhouse
''I'm not sure about the answer to your question, but I believe there are various problems when modeling kinetics compared to thermodynamics. Thermodynamics can for a large part be modeled in terms of equilibria ("stable states") wheras kinetics deals with transitions from one system to the next ("dynamic states"). I believe most of the difficulties that are encountered when modeling kinetics that are not experienced when modeling thermo can somehow be boiled down to this.''

I absolutely agree with what you write. Trouble is that many of those doing the modelling think that they have a complete description based on thermodynamics alone! If you think about it, any first year student could tell you that their description is incomplete, at best.

+ 1 to abge for pointing out the usefulness of computers for calculation and communication, of course he is correct and we are so used to this now that we rather take it for granted.