Planned To Make a Presentation But Notice Serious Errors

[Aero51]

I have an interview with a top company coming up in about 2 weeks. I intended to make a nice slide show reviewing a project that I carried out my senior year. However... I noticed when I was reviewing my data I did not apply certain correction factors to account for my test set up. After applying the correction factors my data looked absolutely pathetic. So pathetic, that the results almost look completely unrealistic. In fact, the "corrected" results suggested that my project would have been completely unfeasible, when in fact it turned out very well. I am not sure what to do at this point. I do not want to go into that interview with no presentation. At the same time, I do not want to present bad or false information. For the record, I am comparing my test results to a computer simulated test.

 

[chiro]

I don't know about the specifics of your project or your field, but I wouldn't deal with false information that you know to be false.

If you need a reason, then one may be that an expert looks at the numbers/assumptions/whatever and realizes that something doesn't "feel" or look right and probes further.

I would suggest you try and find out what went wrong with regard to the simulation assumptions amid other things.

 

[AlephZero]

You could present it as a work in progress.

Describle what you did, show some of the preliminary results, then explain you forgot to include whatever it was, but your first attempt at fixing that made the correlation worse.

The fact that you didn't just accept your first "good result" without reviewing and criticising it should count as a plus point, not a minus.

FWIW when I have an inteviewee talking about their student project, I'm more interested in the process they went through, and how well they communicate what they did, than whether their theory and experiment matched to within 3% or 300%.

If you get lucky, you might even get some ideas about what else to try - most scientists and engineers love solving problems, even if they were really supposed to be conducting an interview!

 

[Andy Resnick]

I have an interview with a top company coming up in about 2 weeks. I intended to make a nice slide show reviewing a project that I carried out my senior year. However... I noticed when I was reviewing my data I did not apply certain correction factors to account for my test set up. After applying the correction factors my data looked absolutely pathetic. So pathetic, that the results almost look completely unrealistic. In fact, the "corrected" results suggested that my project would have been completely unfeasible, when in fact it turned out very well. I am not sure what to do at this point. I do not want to go into that interview with no presentation. At the same time, I do not want to present bad or false information. For the record, I am comparing my test results to a computer simulated test.

There are several issues to consider here, the most pressing centers on exactly what you mean by "After applying the correction factors my [...] results almost look completely unrealistic."

How did you come to be aware of the correction factors? Are the corrected data now so different as to invalidate your original conclusions? Honest reporting of results is the cornerstone of science- if your previous work is now invalid, you *must* correct the record.

Then we have this: "project that I carried out my senior year". Is your advisor aware of your new conclusions? If not, *definitely* speak with him/her ASAP- that person may have used your original results as part of a peer-reviewed publication, a research grant application, etc. and needs to know.

Maybe neither applies to you- senior projects rarely progress beyond a thesis (or final report), and probably everybody moved in different directions afterword. Then it's up to you regarding the thesis- your institution would likely allow you to submit a corrected report if you wanted to submit one.

Now for the slide show. You can't present results you *know* to be false, obviously. If there's not enough time to generate corrected slides, you can always sketch what the slides would have been on a blackboard or whiteboard (Just make sure to request a black/whiteboard prior to the interview!). It wouldn't hurt to point out that while you may *seem* unprepared, in fact you are a true *professional* who cares about their work.

Good luck!

 

[ParticleGrl]

Try and snatch victory from the jaws of defeat- now your presentation is about how we can't trust data just because it LOOKS right. You once did a project where at first blush, the data was amazing, but after applying corrections you discovered...

 

[Aero51]

Thanks everyone so far, let me provide some more context below

1) This was a project I carried out for an extra-curricular team I was a member. I was testing an airfoil I designed and compared the results to a computer code similar to XFOIL

2) I find it strange that my uncorrected curves agree so closely with the XFOIL results. When I applied the correction factors the average percent error was about 50%, without it it was only 10%

3) I did not apply these initial corrections when I presented the results (did not know they existed). No one in the audience (mostly professors) questioned the validity of my data in this regard. This leads me to believe that I might be applying these corrections well, incorrectly. However after reading a paper by Selig, I found I was using the same methods.

4)Contrary to my test results, the aircraft the airfoil was implemented on flew beautifully. In fact, we lost about 1/6 of our wing area due to a faulty take off and it still flew fine. I don't think this was luck. Had the airfoil followed the specs dictated by the corrected results, I doubt it would have gotten off the ground.

 

[TomServo]

Get help from an expert, a prof or ta in aero, make sure there isn't some mistake in how you implemented the corrections.

 

[AlephZero]

2) I find it strange that my uncorrected curves agree so closely with the XFOIL results. When I applied the correction factors the average percent error was about 50%, without it it was only 10%

Unless you have a lot of experience with CFD modelling (which I guess you don't, since you were still a student), maybe the problem is you had unrealistic expectations about how good the correlation should have been, and your expectations were reinforced by the "first time lucky effect".

I don't have any reason to think that XFOIL is not a good implementation of the assumptions built into it, but with any model correlation exercise, never forget that the real world hardware hasn't read the theory manual for the software!

 

[Aero51]

So, I went back and looked at my data and I don't think I had a serious error so much as I came across and interesting phenomena in the flow behavior. I also narrowed down some possible sources of error. Now, who wants to critique my slides? :)