I was speaking to a friend of mine who is about to publish some work. She admitted that she had in her words 'tweaked' the results - not in a major way, but using some 'creative' stats e.g. taking out a few too many outliers etc. she had got a result that meant the work would be 'more interesting and publishable'.

My questions are...

have you done this?
would you consider doing it if your PhD results were non-existent?

In reply to my own questions.

No and No, I don't think so. -my research area is something that could potentially help a large proportion of society, and there is very little on it, so if my research does end up forming the foundations of a new research area, I don't want to be the person that leads everyone up the wrong path!

Never.

I'm a literature student so this wouldn't really apply to me, but what I want to know is WHAT on earth is the point?! It's her PhD, it's not a high school exam which she just needs to pass to get to the next stage. Some of my papers would be a lot more interesting and publishable if I doctored the evidence too.

No I'd never do this either, I'd feel like I was letting myself down if the results I presented weren't really there in the data. With statistics it's very easy to do (I know that if I got rid of just a couple of participants then it'd change a non-significant result into a significant one!) but I definitely wouldn't want to go down that path.

Ha ha, Maria me too. 'This [doctored] evidence suggests that Chaucer may have actually been a woman.'

No, for the reasons every else has mentioned BUT I think it could be a sad sign of the times more than anything as the pressure to publish (and do everything else under the sun) is so intense.

Unfortunately I know of a lot of cases where this happens... outliers removed and unexplained results ignored. I even know of academic reports (that have gone on to influence govt. policies!!) having the images deliberately docotored to enhance the impact of the findings. This is par for the course with some researchers I'm afraid :-s

I personally think it's disgraceful and it is a shame your friend has done it - from my point of view it completely invalidates her research. In simple terms it is cheating!! Not for me!

Put simply- no.

Absolutely not! I couldn't live with myself if I did that!

If you have a look at 'The Burt Affair' on wikipedia it details reserach where a respected academic made up a whole heap of research - even fabricated the research assistants that were supposed to have worked with him!!!

It is obviously an extreme example but unfortunately there always have been and always will be academics who work without integrity

Whilst not an advocate of outright cheating or stats faking, I think that to an extent a certain degree of selection is what everybody does when they present their work. Any results section of a thesis or paper will present only some of the story or a partial story because conclusions need to be drawn. Also, deft use of language or highlighting certain sources of figures can "frame" a certain "fact" - or lead the reader to draw specific conclusions. We all do this, our theses are not factual or objective in any respect.

Yes, as a historian I have not falsified sources and have made sure that arguments make historical sense, but so much sub-conscious and conscious sifting and re-ordering takes place.

So I wouldn't be too hard on the person in the original post,

======= Date Modified 17 Feb 2010 11:58:02 =======
======= Date Modified 17 Feb 2010 11:57:42 =======
I did see a programme the other day on an archeologist who claimed to have found the 'missing link' - this was back before good dating techniques I assume. He made loads of cash I think and kept on 'finding' these things. They later found out that the skull he found was quite 'young' and definitely not cave man era, all the 'findings' were also in a large circle around his home!

http://www.bbc.co.uk/history/archaeology/excavations_techniques/piltdown_man_01.shtml

I see most people who have responded so far are clearly saying no, but I think that often it is a fine line between e.g. "cleaning" up a dataset, so that no erroneous outliers f*ck up your stats and "tweaking" results overall by selecting the data points that fit into one's interpretation. (The latter being not okay, whereas the former could be justified). I think it always depends on how procedures are described in papers, and e.g. if your friend, Sneaks, describes how many outliers were removed, what procedure (even if it was eyeballing it) was used, then, it is open and transparent, which is what is needed (and reviewers can criticise and judge). Yes, ideally, you would also show the stats including those outliers (in an appendix for instance).
So for your questions, yes, I have e.g. removed outliers (following a procedure) and described it in my paper. But yes, I see your point, it is of supreme importance to maintain academic integrity, and e.g. fabricating data is not acceptable!! I always ask myself whether my data "manipulation" changes the overall conclusions from the work, and if so, then something is not quite right, but if it e.g. strengthens a case by making things clearer, then that could be acceptable (and I think that approach is widespread in science..)

I did contemplate it, I did some preliminary analysis and had a nice little significant result after 100 particpants. THen another stupid annoying 100 participants did my survey and it turned very non-significant - stupid participants! :-s I did think of just saying I only had a sample of 100 but I just wouldn't dare - I would assume someone one day would want the original data or something and it would all come crumbling down like a house of cards around me!

This is a pet hate of mine. If you modify the data set then you are changing the original experiment and any statistics lose their significance. There are certain measures you can use which minimise the effect of outliers but removing them entirely (unless there is real cause, eg that was the petri dish accidentally left on the radiator for an hour) is as good as making up your results, statistically. Of course when discussing results, in a paper or thesis, one focuses on the interesting ones or makes things seem more important than they are, with words, or maybe uses certain measures to get the best representation of the data. This is not the same as claiming statistical significance when it's not there. Completely understand why people do it though.