Speaking at conferences with partially incorrect results

[hunt_mat]

Hi,

I am due to give a half an hour talk at a conference. I have some analytical and some numerical results that I want show. For one part of the work the numerics works fine but there is another part of the work where the general shape (I an computing waves in interfaces) is correct but the but the values at the end points are incorrect.

My quandary is this, do I explain the method, show the results and explain why the results are wrong or do I leave that part out of the talk entirely?

Not too sure if this is the correct forum.

 

[twofish-quant]

I'd include the results and explain why you think they are incorrect. One reason for presenting that is that you may help someone else that is seeing the same thing, and that would make for a useful Q&A.

 

[hunt_mat]

That was my thoughts too but I wanted other peoples point of view.

 

[Andy Resnick]

Hi,

I am due to give a half an hour talk at a conference. I have some analytical and some numerical results that I want show. For one part of the work the numerics works fine but there is another part of the work where the general shape (I an computing waves in interfaces) is correct but the but the values at the end points are incorrect.

My quandary is this, do I explain the method, show the results and explain why the results are wrong or do I leave that part out of the talk entirely?

Not too sure if this is the correct forum.

Absolutely present *all* the results, especially if you understand *why* the "values at the end points are incorrect" (or even if you have a few good guesses). Other people may be interested in knowing the appropriate limits of your simulation. Never try to conceal something- someone may ask you about those results, and then you are stuck.

Clearly, emphasize the better results, but being able to say (for example) "using this method does not work over here, most likely because of [...], but we think we can extend the region of validity by [...]" is a good thing.

 

[hunt_mat]

The method should work, this is PhD work I am doing. The field is fluid dynamics, so the numerics should be bang on, the fact that I am not getting the correct boundary values is somewhat embarrassing. I tried to talk to my supervisor but he isn't answering my e-mails.

 

[AlephZero]

It probably depends what sort of conference it is, but I think there is a difference between "these results seem to be wrong but I don't understand why" (i.e. they don't match boundary conditions when they should, they don't match experimental data, etc) and
"I know these results are wrong, becaose I made a mistake".

I don't think anybody would have a problem with the first scenario, but there isn't any value in presenting the second. If you do get into that situation, just be honest and say some of your results you intended to present are missing because you discovered an error, but you haven't had time to rerun them before the conference.

Presenting resultswhen you know why they are wrong is at best wastiing everybody's time, and at worst may confuse the message you want to get across about your correct results.

You may be able to include the corrected results in the conference procedings, even though you couldn't present them "live".

 

[hunt_mat]

As I said, I think that the general shape is correct but that it should be shifted up by a given value.

 

[Sybren]

If it is clearly not correct yet, you can leave your current result out and mention the approach you are using in your outlook section, this might get you some audience response which can help you solve it later on.
Otherwise, you definitely need to talk to your supervisor about this. Don't send emails, calling is the quickest way to get some feedback.

 

[hunt_mat]

My supervisor is notoriously bad for e-mailing and I don't have his number. hence the request for advice here.