Submitting this month, just been told a better method to do analysis. Aargh! Any advice?

[LauraAshleyDuvetCover]

I presume I can't submit and then submit a pre-emptive correction before my viva can I?I will obviously talk to my supervisor about this, but just gathering a few suggestions first.

Throughout my PhD I've sent samples off to another university, had the results back and analysed them myself. It's been fine, maybe not quite as elegant (and certainly takes me longer!) as somebody who does a PhD based around this analysis method, but no real problems.

I submitted a first draft of a chapter and my supervisor asked me to include more detail on analysis for one set of samples. I'd only put in a few data sets because the preparation technique didn't work very well and I found something better. At the time I looked through the data sets, but didn't analyse them all because I could see that what I'd hoped for hadn't worked (big mistake!). So now after analysing them all how I usually would, and finding it doesn't work very well (I know I should have realised this earlier) I've e-mailed the person who ran them, asking how to solve a particular problem I was having. They had a look and said "That data will be difficult to handle, and you won't really be able to do , you really need ". I have never done and there's no way I can learn it, do it, and write it up (as well the other last minute tweaks I need to do) in a week. gives me some data but it's more indicative than reliable, and doesn't work very well for some samples. will give me higher quality data and for every sample, although it still isn't a magic bullet — the data is still complicated.

I feel like my options are

1. Write up what I have done and acknowledge it isn't very good. I have no real explanation for not finding out about and doing earlier.

a. Put technique in future work and presume I'll get it as corrections. b. Write it up, submit it. In the gap between submission and my viva, learn , do it, and either send the section to the examiners, or turn up to the viva with it.

1. Find somebody who does and throw myself on their mercy and ask them to physically sit with me and help me.

I think it's too late for 2. I feel useless that I didn't realise I needed this technique earlier, and also stupid for not taking it to my supervisor earlier (although in slight mitigation, they had seen it both as raw data and in the thesis and had never mentioned this technique. If I'd gone to them saying that I was really struggling though they might have thought of it). I've already had a generous extension because of covid/lack of access to resources/difficulty of writing during the pandemic, and "didn't know how to analyse her data properly" doesn't seem a very good reason to ask for more. I'm also getting a bit fed up and just want it to go really...

Any wise words MNers?

It's hard to follow without knowing the methods... I'm imagining you used linear regression when you should have used negative binomial or Poisson, for instance.

My big fear is the examiners will just flag your work as needing Analysis Methods (2) anyway. You won't save yourself work by avoiding the more suitable stats method, you'll give yourself extra work by preparing a lot of results tables that are thrown out after all. You'll just have more corrections to actually get the PhD, in the end.

Correct analysis of the data is just too fundamental to be overlooked (probably).

Is this stats / computational analysis or a lab analysis method?

If it’s a lab assay you should have done, it’s too late now and I would do (a) and hope for the best

If it’s stats, I would do it for the viva

It's computational analysis. I'll explain a properly (and hope none of you recognise me!).

It's powder X-ray diffraction data, and has a high background because the sample is wet (there's nothing I can do about that, it decomposes if it gets too dry). I've modelled the water as part of the background. I'm then refining it to look at different phases that are there depending on the preparation method, but I'm struggling to get it to fit well. I now know that there's a way I could have done it where I could have mathematically modelled the water as an amorphous phase, which then takes away a lot of the mathematical complexity of the fit because it's easier to fit x+1 phases and a flat background than x phases and a complex background.

I've read up on the technique now, and I had seen it mentioned before in my general reading about PXRD, but it seems to mainly be used when you have a crystalline and a glass phase and you want to quantify them. I just didn't recognise it as something that would have worked for my samples.

If you have other good results, I would write up what you have and do this analysis before your viva, and send the re written chapter to your examiners before then or show it to them then. You could include both methods in your chapter.

So does this analysis technique refer only to the preparation method that you excluded early on because it didn't work very well or is it a technique that you should have been using for all your analysis? i.e., is it possible to say that your analysis method was good enough to tell you that there was no point proceeding and you did not need to use this other analysis technique to confirm it?

If it is the first, you could just retcon it as using a good enough technique for a first pass analysis to see if it was worth continuing.

If it is the latter and you could / should have used the technique for a significant part of the analysis of your PhD, then I guess it depends how well-known and accepted this technique is - would people in the field generally have expected you to use this technique and it will be considered actually wrong that you did not use it? If this is the case, go for option 1b, do the analysis for the viva and work on an explanation of how using this technique would not have fundamentally changed anything, or, preferably, delay your submission. I am not a scientist, but in my area, if someone had used an actually wrong analysis method, knew this and had not corrected it, they would be on the border of major corrections and fail so this approach is risky and could see your PhD dragging on for months after the viva.

Or is it more a case of so-and-so has come up with a new technique for dealing with this issue and in the future it would be worth testing it further. In which case still go for a combination of option 1a and 1b. Acknowledge the existence of this technique and for the viva have some analysis with this new technique and be ready discuss how it might be better.

If you're up against it with a hard deadline I would write it up based on what you've done. Then in your Discussion mention the limitations of the approach you've used and say that an alternative method that might have produced better results is method b. The examiners may ask you to do b as a correction (and if you're subsequently writing up for publication you should try b) but you will have more time to run b properly at that point instead of trying to rush it now.

If you can do 1b, go for that. It will be useful if you ever want to publish the work.

But really, you should be asking your supervisor not Mumsnet. S/he will know the norms of your field & the best way to present stuff

How crucial is the XRD data? I would be tempted to present as little as possible in the thesis now and maybe include one sample (one of the better ones) where you do both methods and compare results. Could your collaborator help with one sample this week? You can say future work will be doing analysis using method b and include it at the time of the viva, but if it's not great data anyway and the analysis will have to be redone I would just accept you'll have this as a correction and try to waste as little time on analysis with method a that you won't use for the final thesis or publication.

Also, being on your own learning these sorts of techniques is tough and usually school of hard knocks. You have my sympathies!!

Quick update.

I've spoken to my supervisor, and we chatted it through, and they said that wasn't great, would be overkill and to do simpler technique instead (which I had done for some samples, to help with ) and talk about why doesn't really work. If I'm asked about in the viva, I can talk about why it might work and why it might not (my supervisor isn't actually convinced it would, because it isn't really what it's designed to do).

So we've extended for a month. I think it probably suits us both — supervisor is head of Covid response, so it gives them more time to read it through as well. And I might actually manage another in person meeting with them! I want to get it in asap, but feel the pressure is off a bit now. I have a plan and some wriggle room on the timing!

Thank you all — it meant I had my ideas in order when I went to speak to them. And particular thanks to badg3r for saying it's hard learning it on your own. It really is! I'm definitely much happier in the lab where I can 'see' what's happening than on a computer where the software is a bit of a black box.