Hi all

I just wondered if anyone has experience of something like this... the stats method I want to use (should use given my data structure) is quite complex. Because of this my main sups tried to get another sup on board to supervise the stats part. Anyway, turns out the new sup isn't an expert in the method either, and we're going to sort of be "working it out together". Basically, I am quite concerned as I want to feel certain that the models I come up with are good. The literature is helpful in understanding the method and when it can be applied, but the practical issues of building the model aren't greatly covered (for what I specifically want to do). Is it just luck if you happen to know someone in the field of statistics (or who has expertise in a particular statistical method) who might be willing to take a look at what you've come up with?!

Tudor

I think there is an element of luck in having someone with the right experience to hand, yes. I am in a similar position, with the stats not being quite the straightforward ones I'd hoped to use, and everyone in the dept has been asked for advice. At the end of the day, I'm doing an analysis that feels to me to be the best fit for my data, my supervisors are on board with it, and they've advised me to write my decisions/justification into the chapter, and to be prepared to answer questions about alternative approaches at viva.

Thanks Chickpea. It does sound similar. I'm happy and keen to learn the approach - I've really enjoyed understanding why it is suitable for my data etc - it was me who realised I needed to use it and sort of had to persuade one of my sups about why it was necessary. So I have no qualms about defending the approach and writing a strong justification for it vs alternatives. I just feel a bit worried I suppose, having met with the new sup who was specifically asked to join for support with this, and then he says in the meeting that he actually isn't 100% sure of whether we've done it right (not whether it is the appropriate method but whether we're actually doing it right in the software), and why don't I just do a simpler approach (even though the simpler approach doesn't reflect the data structure and would violate a major assumption!). I was a bit shocked I guess. And I'm wondering who I could talk to. I guess it is just luck if I get to talk to someone in my department who knows someone else who is more skilled in using the approach...

It sounds like you have a really good handle on why the technique is suitable, but just need some help implementing it. That said, if you plan to get a statistician involved, I'd do it at an early stage as is possible. I've been the person people have come to at the end of a process to 'check their work' and it's very frustrating on both sides if something was flawed at the outset and the whole thing is wrong. I don't think you'd fall into that trap as you're taking a considered approach, but as a general rule I'd say the sooner the better for statistician involvement.

Does your department/faculty have any kind of central data/analysis support unit. Some places have regular drop in sessions for scenarios like this.

Thanks Hazy Jane! I've just found out that we can get statistical advice from the biostats dept! I have to send an email detailing my scenario and what I need the support with and they should hook me up for a session with someone.

Now I have another question... I guess the best way to be with my supervisor is transparent about what I want to do... (go talk to someone else). I am now worried that he will not take it well that I want to chat to someone else. Any advice on that front?

I got some stats advice during my Phd. I think I just said to my supervisor that's I had the opportunity to speak to a statistician so I had done it. I never looked at it like I was going because they didn't know what they were doing, I viewed it as being fur my benefit. Maybe sell it these terms?

I don't think they should mind you speaking to someone with the right expertise - it shows you're being proactive and careful.

Thank you both. I will do it and sell it on those terms. : ) Now let's hope there's someone available to talk to me! One worry at a time lol! Have a great weekend!

I paid an academic statistician £20 over payperhour website and got advice and got it checked. It was the best thing I did instead of trying to get advice from a statistician in my department/uni. I'd be happy to send you details if you like.

Hi Hugh, yes please. It would be good to know the details for peace of mind and in case there is no one available with the expertise at uni.

Find a summer school that specialises in that methodological approach. No one in my department really understands my methodology (except my supervisor who has a familiarity with it) so I will be attending a summer school to brush up my skills.

Thanks Teaddict. It's more a matter of someone checking the R code really. I've attended workshops, read and done so many tutorials etc that I'm stuffed with knowledge on the approach. Just need a person who actually uses it and is skilled enough to be able to map their statistical knowledge on to a different research area to say, "yes, what you've done is right", or "no, you need to change that, otherwise...". Hoping I'll find someone at uni at the drop in thing. Otherwise, asking an external statistician would basically achieve the same function but at a cost.