Statistical Science


Transcript: Episode 3

Terry Soo  0:17  
Hi Claire, How are you doing? So thank you for coming out tonight to speak with us.

Clair Barnes  0:24  
Fortunately, I didn't have far to go, you know, from my sofa to here. It's about 10 foot. So that's come and speak to you.

Terry Soo  0:32  
So it's been about one year now that we've been stuck at home, I guess. Any thoughts on being stuck at home for a whole year?

Clair Barnes  0:40  
Well, to be honest, because I was doing my PhD mostly working from home before that. It's been way longer than a year. Yeah, I feel like this has been about the last five years really, although probably everyone feels like it's been the last five years.

Terry Soo  0:54  
What have you been working on your PhD?

Clair Barnes  0:57  
I'm looking at weather forecasts doing a little bit of postprocessing work and some uncertainty quantification. So we've got a Bayesian framework that kind of models the different elements that go into your forecasts that your your operational centres produce. And we're looking at how we can use that to try and post processor, which basically just means correcting the biases and trying to get a decent assessment of the uncertainty that is in your forecast, getting them well calibrated.

Terry Soo  1:23  
So what does this do? Does this tell me what the weather is gonna be tomorrow? 

Clair Barnes  1:26  
Well, you already have a guess of what the weather's going to be tomorrow, what we need to do is correct. That guess basically, so all of the models have some biases in them. And all of these different operational forecasting centres like the metaphyseal, majia, France, and every country has its own weather forecasting centre, and they all have their own models. And they all say their model is right. And all of the models disagree. So what you can do is you can combine the models into a multi model ensemble, and see if you get a better forecast. But actually, because of the way the models are constructed, so they have kind of shared code, and they have the same sort of assumptions about how the atmosphere works. So they all have kind of the same bias generally, or the same kind of errors in there. Just adding more ensembles, you're not getting closer to the truth, you're getting closer to what all the ensembles are modelling. And then what you actually want to do is get to over here, so you need some sort of post processing correction on all of your ensembles to try and get closer to what you're actually trying to forecast. So

Terry Soo  2:23  
These models do better than like, you know, I don't know if I'm a farmer in the 1855 or something like that. I look at this thing called the Almanack or something like this, right? And it says, like, march 23, it rain? So last year, read this year, maybe it's all the same time. All right, again, right, you know, that I know, what they used to do in the past, it gets it does better than this. Right?

I really hope it does better than that. I mean, you'd have to show me the book, we'd have to do some proper, you know, statistical analysis, some sort of rigorous investigation, but I my hunch would be yes, we would do better than that.

So what is it? Is it based on? Like, it's pretty basic, up to date information, right? Cuz if I go to my cell phone, and it says, it's gonna rain, sometimes it just changes, maybe the indoor some cod moves or something?

Clair Barnes  3:06  
Yeah, precipitation is hard to forecast. That's the tricky one. So I'm mostly looking at temperatures because they're nice and smoothly varying. And they're sort of a bit Gaussian, whereas precipitation, you've got this kind of weird mixture of, does it rain, does it not rain? And then when it does rain? What what does that look like? Like? What's the distribution of the rain? Amount? It's really hard to model when you start trying to get into rainfall modelling. It's very complicated. So I'm mostly I'm looking at just temperature forecast, post processing at the moment to answer your actual question, which I didn't do. Yeah. So all of the models, they get updated based on the latest information. But missing information can make a really big difference. So they're probably the app on your phone is updating current forecast as quickly as it can. But if they run a simulation, and it outputs every hour, then you might get an update on that hour. That changes something quite big. So I was talking to some meteorologists in Switzerland. And there was a case a few years ago, where there was a massive flash flood in one of the big valleys in Switzerland, you know, people drowned. And it wasn't, I think it was on Christmas Day, actually. And, you know, houses were damaged. And it was it was a massive event. And there was no warning whatsoever. And it because was because there's one radio song, which is like a balloon with measuring equipment on essentially, they hadn't got the data back from that in time to include it in that simulation, forecast simulation. And when they went back and ran it with that radiosonde in there. There was this huge storm was was clearly there, and it was clearly forecast. So if they'd had that one piece of information that came in like an hour too late, they could have warned everyone. So it really can be that kind of sensitive to the data that you have available and in the really short term.

Terry Soo  4:46  
And in the end, but the temperature isn't like this though.

Clair Barnes  4:48  
The temperature less, I mean, if it rains, it's going to change the temperature but generally temperature doesn't change by a huge amount, unless you've got a big weather system coming in. And normally you can see those things well in advance, because most well, especially at the Sun view, more so in winter, but kind of at this time of year, a lot of it, a lot of our weather is dominated by kind of frontal, big synoptic. You know, when you watch the weather, and they have these big kind of isobars swirling across the whole country that kind of dominates our weather.

Terry Soo  5:16  
So when you're always interested in this is what brought you to UCL the weather or...?

Clair Barnes  5:21  
No, I didn't want to do weather, I wanted to do climate. So the same thing, and not the same thing. No weather is like what's actually happening day to day and climate is the kind of longer term average behaviour of the weather. So if you kind of average the last 10 years of weather, that's your climate. What's happening tomorrow is the weather. Yeah, so no, I can't UCL because I read a paper of Richard's, that I kind of said, Oh, this looks really interesting. And it's the same framework, but about climate. He kind of proposed it. Eight years ago, I think. And I saw this paper and said, Oh, this is really interesting. You could do sequential forecasting, you do this, this this, would you be interested in supervising a PhD, and he wrote back to me and said, someone's just done a PhD on that. But you could see if it works for weather was was kind of came out. So that's what I ended up doing. But now I'm doing a postdoc that is looking at climate. So I'm kind of back where I originally intended to go.

Terry Soo  6:19  
I see I see I see. So when you read this paper, were you? What were you doing when you were your Master's student? Or were you working in industry? What were you doing when you read that?

Clair Barnes  6:28  
I was, what was I doing at that point? I think I was between things. So I did a maths, I was working. And I decided I needed to retrain because I didn't have any qualifications that let me do the stuff I wanted to do, which is basically playing with numbers. So I went and did an open university degree, because my original written degrees English literature, alright. See, obviously, that not great for employability when it turns out that your brain is very much an analytical numbers brain. So yeah, I wouldn't didn't open university maths and stats degree, I quit my job and went and did a master's at Warwick. And I wasn't really thinking of doing a PhD until I did my master's project and really enjoyed it. But by that time, it was June and it was kind of too late to apply for that tranche of these. So I had a year of not quite dead time. But at the time, I thought I had a year of kind of just waiting until I could start a PhD. So I was just doing kind of temporary jobs. And I actually did a, I always referred to it as my pre doc up at work, working never kind of sick because it was it was replacing a postdoc that had got another job, but it was only for the last six months of the project. And that was looking at X ray imaging. So that was what I was doing.

Terry Soo  7:38  
X ray imaging. Ah, I see

Clair Barnes  7:40  
Looking at the development of defects in X ray detectors. It's really cool.

Terry Soo  7:45  
So what So what about that wasn't your master project, but what was what was your master project,

Clair Barnes  7:49  
my master's project was statistics in Anglo Saxon archaeology. I like that it kind of tied back to my English degree because I did lots of old English, Anglo Saxon archaeology, in that I was kind of a mediaeval and previous specialist.

Terry Soo  8:04  
Is that how you chose that topic, because you had the experience?

Clair Barnes  8:08  
I'm not sure experience is the right word. But it certainly appealed because it kind of it had a nice sort of symmetry to it, I guess. And also, when I spoke to the supervisor, it just sounded like a really interesting project, because I quite like applying statistics in areas that aren't necessarily well trodden. If that makes sense. There are always lots of projects in like financial statistics and medical statistics, but it feels like everyone's doing that. And this was just just this little weird kind of niche thing of looking at whether there was any evidence in these Anglo Saxon sites that that there was sort of a greater plan than someone just rocking up on a hillside and going, I'm going to build my house here, because it's a nice view. And then someone else walking up and going, I'm gonna build my house there, because that's a nice view. And it's far enough away from you. So it was looking at the angles between walls to see if you could kind of identify evidence of common planning between different structures. Sounds kind of cool. I got to do lots of stuff with directional statistics and lots of image processing, because all of this stuff came from all of the data that I had came from scans of maps from archaeology books. Oh, and speaking of archaeology books, this is this is my favourite thing. This has my master's work in it. Wow. I'm in the Appendix.

Terry Soo  9:20  
Thats awesome.

Clair Barnes  9:21  
I know. See, the reason I didn't apply to the reason I didn't English degree was because growing up, I always wanted to be a writer. And I'm like, I'm gonna bug. Not quite how I plan but you know, it worked.

Terry Soo  9:33  
So I guess if the houses were all random angles, then there'd be no planning. But if they're all sort of pointing.

Clair Barnes  9:39  
Yeah, there's evidence that they're all kind of aligned, then that suggests a greater plan. Because this is all from looking at evidence from the period that's often been known as the Dark Ages when there was kind of this assumption that we're all just kind of heathen barbarians jumping around the fields, because the Romans weren't there telling us what to do. We didn't really know what to do with that actually, there's there's a team of Oxford who wrote that book, who believe there's evidence of kind of more structured planning going on, I'm not going to try and remember which areas of the UK they were looking at specifically, but they had a couple of sites. And they wanted to look at some evidence of that. So my supervisor had looked at, they're also looking at whether there was evidence of a common unit of measurement being used, which is all approach. So you know, where all the buildings to within a certain distance, because all of this information comes from postholes, which is, so you basically got kind of holes where fence posts were stuck in the ground and make a move over time. So you can't, it's not super precise. And there might be some slippage. And it's really interesting. It's really interesting. And it kind of appeal because I can remember doing an archaeology module when I did my English degree. And you people would make statements like, Oh, yes, most of the graves are aligned in a north south direction. And I remember at the time going, I mean, most of them, how aligned how north south? Does that mean? And I think there's, there's a lot of stuff you could do with statistics. Okay. Is that actually true?

Terry Soo  11:05  
All right. It might not be I guess, right. It may not be right. This is

Clair Barnes  11:09  
It's just someone eyeballing it and going well, I think they're probably all in a straight line. I don't know, but there are things that you could do to check.

Terry Soo  11:17  
Yeah, it's probably something Harada has told us and then we just believe it. Right? So it sounds like, you know, statistics and science have a lot more to offer archaeology than just like carbon dating or something like that? I don't know.

Clair Barnes  11:29  
I think so. But then I am someone that I like, my evidence to be backed up by proper evidence, rather than by someone going, I think this is the case and everyone going, Okay.

Terry Soo  11:39  
I'm not sure carbon dating even applies. Actually, to something like this.

Clair Barnes  11:44  
Not with looking at the direction of maybe because you need to date your structures. So you can only see if you've got common planning between structures from the same time. Ah, so carbon dating, you would need to date artefacts that are found in the same area to establish well not explicit. That's that's largely speculating, I'm not an archaeologist, there's probably archaeologists, if an archaeologist ever saw this would just be getting know what you're talking about. So I should step back from pronouncing.

Terry Soo  12:14  
Probably not gonna watch this. So you're probably safe, right.

Clair Barnes  12:17  
I'm kind of assuming so but I just don't want that to be one furious archaeologist going, you've missed representatives.

Terry Soo  12:23  
So who was your, who was your advisor at work?

Clair Barnes  12:26  
It was called Wilfred Kendall, who comes from a noble line of statisticians. He's not Kendall of Kendall's Tao. There's a lot of Kendall about all that a different candle. That's a different Kendall, and David Kendall who I think set up one of the statistics labs at Cambridge, or the statistics lab, at Cambridge, I'm not sure of the name of it. And he actually did a lot of the work on on this common unit of measurement. I think the second set up the method that was used are to establish what there was a common unit of measurement, and also looked at evidence for ley lines, which I love. What are the geometry of ley lines? Ley Lines are mythical lines of power that supposedly, like sacred sites are aligned along and they're supposed to either conduits for mystical energy or something? A bit woowoo. But, you know, he actually took some statistical methods and was trying to establish whether there was any evidence to support the fact that they might exist.

Terry Soo  13:23  
Oh, wow. Do they?

Clair Barnes  13:26  
I don't remember reading a conclusive decision on that. But I don't I don't think so. I get anyway.

Terry Soo  13:35  
So you mentioned that your previous with the way you guys call it here in the UK course of study or degree, we just say degree, you mentioned your previous degree was in English literature, have you thought about? So you know, one of these things? When I first got exposed to stats, things that I heard people study was how many words they I don't know why someone would want to know this, but I guess it is kind of just see how many words the Shakespeare know. And then they went through his things is like, wow, you know, in this play, there's so many unique words.

Lot of words, I think that's one of the things that people always say when, because I always get a bit kind of Shakespeare. I mean, I like Shakespeare, like Shakespeare less than I did after I was forced to do a six hour exam on Shakespeare. I do I kind of object to this dividing of literature into Shakespeare and everything else i which is kind of how its taught or how it was taught when I was at school anyway. But then one of the arguments against that is actually we have a lot of words that Shakespeare either invented or wrote down for the first time. I can't say for sure that he invented them. So that that is true. Yeah, I remember spending a lot of time in the UCL library because I did my English degree at UCL, they have these concordances, which is just a book of all the words in Shakespeare, like you can get concordances a lot of things but it's basically a book of the words. Not necessarily in the right order, obviously. Yeah,

I guess you're trying to estimate this guy. How many unique words this guy knew? And you can never know

Clair Barnes  15:04  
Yeah, I don't really know what you would do with a concordance. I just remember kind of stumbling across them boggling slightly.

Terry Soo  15:10  
So what's the thing? It's a book of all the words that Shakespeare use it? That's called the concordance. 

Clair Barnes  15:14  
Yeah, but not necessarily of Shakespeare, you can get a concordance to the Bible, you can get a concrete, it's like the dictionary of a particular body of work. 

Terry Soo  15:22  
And people compile this, like going to get help for the data stuff or something.

Clair Barnes  15:26  
I guess it's like Google Trends. It's not quite Google Trends, you can search for how often a word is used on Google. I think.

Terry Soo  15:34  
 Okay, okay.

Clair Barnes  15:35  
I think so. Because you can always track like, I think the big one that people have done recently is like, searches for Brexit and things. But you can also check how many times words appear in books. So you can see the track the usage of words over time.

Terry Soo  15:48  
And wonder in the past, that was someone's PhD thesis.

Clair Barnes  15:51  
Yeah. Yeah. So back from the time when you would read statistics papers, and they would thank the people at the computing lab for letting them use their computer and for letting them run 100 Monte Carlo simulations because they could afford in the year that they had. It's kind of parallel to that, I think.

Terry Soo  16:13  
That's good. It's got plenty right? Did you ever think back to what we do now? Right? When we look for paper, so what when you look for paper, so I confess, I'm more of a math guy. So I just go math sign that and most of the math journals are indexed in our Senate. And in the past, I guess these guys would have walked the library started browsing, or they would just asked a professor Kendall, you know, do you know where this paper would be? And then it will probably turn around to this shelf and go, Oh, I've got this book here from from my dad era, it's in there probably that's probably what he would do. He probably knows all the literature. So what what do you do when you do research? What do you what do you browse? Or most of the things indexed by math sign that or how how's it working?

Clair Barnes  16:50  
I use Google Scholar, I don't tend to use my sign up. Because I have that kind of meteorology, climate stats kind of intersection to deal with, I can't really just go to one place. So I find that Google Scholar is it kind of just sounds everything

Terry Soo  17:07  
Yeah, that's right. Oh, I see. Because you actually have to recite you have to actually read science journals. Possibly, right?

Clair Barnes  17:13  
Yeah. About the actual, like, the physics behind it and stuff. Yeah.

Terry Soo  17:18  
Is that Is that hard? Because I find that even hard to like, if you told me to read like this different area of probability, I find that you don't even know I can't even read.

Clair Barnes  17:27  
Probability is hard. Me read a probability paper I'd just back away slowly; just gonna ask Terry what it says I probably. Yeah, no, but I mean, that's something that I'm doing quite a lot at the moment with the with this postdoc that I'm doing. So we're looking at climate projections, and we need to compute a lot of indices of these climate projections. So the first thing we needed to do is okay, what are we going to compute. So we want indices of extreme weather. You can't just say, Oh, we're going to compute compute maximum of this minimum of this, you need some rationale for why that's a useful thing to compute. So I have spent weeks just trawling for papers, go write drought indices, I need to learn drought indices now, right? What? What can we do? What's likely to result in flooding? What's the kind of most useful metric that we can use to summarise the climate data? Yeah, so I've had a big old crash course in kind of weather indices and impact relevant metrics. It's been really interesting. Well, that's one I like that, because you get to nosy around and other people's research events and be like, I'm just gonna read all the good papers you've written and then do some stats on it. Okay.

Terry Soo  18:33  
So when you do stats on it, did you do come up with a model? Or what what did sort of like in general?

Clair Barnes  18:41  
In what I'm doing at the moment, it's very much a descriptive phase of the project. So the project only started kind of in earnest at the beginning of this year. And honestly, it's taken me three months just to get the data to the right shape is climate data. Not we're not discussing that. I'll just go from that will be. Yeah, so the kind of first step is, because what we're actually doing is we're comparing an ensemble of projections that the Met Office have come up with, which uses one global climate model and one regional climate models kind of predict the UK climate for the next 100 years. But they've kind of perturbed the physics. So they've got 13 Different realisations of it. But it's all using one model, ultimately. And then there's this Codex ensemble that has lots of different combinations of global climate model and regional climate model over the same area. So that gives you a lot more about the model uncertainty. So at the moment, what I'm doing is kind of comparing. Here's what the UK ensemble says. And here's what this Codex ensemble says. And that's kind of the first phase is just describing it and saying okay, well the metaphysis ensemble explores the kind of warmer drier end of the kind of what the projection were all of these projections are saying so if you kind of supplement that with the Codex ensemble, you get a wider spread of You can't really treat it as a probability distribution on this because it's not calibrated on whatever quantity you're looking at. It's not a distribution that you get, you get different scenarios, basically. So the Met Office ensemble is kind of showing warm dry scenarios and the euro Codex ensemble, we think is going to be kind of cooler, wetter scenarios, because that obviously has a big difference on like, what your global mean temperature is going to do and what your changing risks are.

Terry Soo  20:28  
Are you doing a critique of these models, or are you combining the models?

Clair Barnes  20:31  
Critique makes it sound like it's of the models themselves? It's it's going to be reporting on the kind of differences in the models where they complement each other. So if you're a researcher that is interested in looking at drought, and you want to know what's going to happen, if you want to know what's kind of the least severe drought, you don't want to be looking at the Met Office. This is massively oversimplified, by the way.

Terry Soo  20:57  
That's what you need to do for a probabilist. Right, this is, this is fun. This is what you need to do for probabilists. 

Clair Barnes  21:03  
That's good. I can I can do massively over simplified. So yeah, if you were if you wanted to look at drought indices, so if you only look at the Met Office, which is known to have this warm dry bias, then you're only going to get the kind of more severe end of what's projected. So you really need to know kind of what your spread is within each of these indices within each of these ensembles. So that you know which one to pick for what you want to consider, kind of what you want to evaluate.

Terry Soo  21:28  
So I guess like, correcting the bias of these things, is gonna be harder than dividing by one over n minus one instead of one over n that I guess, right?

Clair Barnes  21:34  
We're not getting a bias correction yet. So this is the first stage this is the kind of descriptive stage is literally just going to be making a lot of pretty plots and going, Oh, look, this one's work. This one's dry. This ensemble shows this, this ensemble shows this. And it's it's going to be that kind of descriptive analysis. One of the things we're going to look at which is introducing this framework of Richard's, that I was talking about with the kind of uncertainty quantification and that combining models and trying to do some bias correction, is we want to look at the uncertainty, the sources of uncertainty. So how much of your model uncertainty comes from your choice of global climate model? How much from your regional climate model? Can you split out in a meaningful way? These sources of uncertainty and can you use that to design future experiments in a more efficient way? Can you do anything about it? Can you can you bias Correct? Can you bias correct for 20 different variables simultaneously over the whole globe?

Terry Soo  22:28  
But I guess, just from some naive standpoint, if you know there's a bias surely shouldn't be the correct for right?

Clair Barnes  22:35  
Yeah, that is the naive way of doing it. Unfortunately, I wish that was the case. I really wish I had that simple, right? This if you correct your temperature if you change your temperature by two degrees, because your model is two degrees too cold. So yeah, up your temperature by two degrees. What's that gonna do to your snowfall? Well, second due to precipitation, changes whether it rains and it changes it. So all these physical dependencies, so you had to be really careful with statistical bias correction. Because you you just break the physics basically, you can you can make whether that couldn't happen.

Terry Soo  23:12  
It's like when I played with, my brother took a music course he came home with this programme called finale. And then I loaded on my computer, so I can read music. I just got to put these you know, in Chinese we call them black beans. I think maybe it's only Cantonees people call these black bean sauce. Other put the black beans in. And then I started playing this tape. And then he he wrote some pieces first class, and then the professor. Most people in the class had, were classically trained. We didn't grow up playing anything. And the professor's like, yeah, no, this is this is fine. But you know, this is not human.

But Richard, the other day mentioned that, oh, you're joining us on the call. And I was also Claire, you're you're a postdoc right now?

Yes. Even though I haven't quite submitted my PhD yet.

How's it feel to have moved on to the next level?

I didn't really feel like I have yet. I don't think I can claim that I've moved on until I've submitted my thesis, which I really need to crack on.

But that's already all done, I guess, right? Because you're working on this new stuff already.

Clair Barnes  24:14  
It's, well, that's the problem. It's not quite finished writing up yet. And now finding the time to finish writing it up is not easy. It's getting that it's getting that went up. Yeah, no, it's um, it's nice. It's kind of odd not being able to go into the department. Actually, I think it was a bigger change. Just whenever I went into lockdown and everything went online. And suddenly things that I would have had to go into the department for like the departments and seminars and all this kind of thing. Suddenly, it was all online. I was like, Oh, I can actually join this without having to commute which when I don't have an office is not super motivating to you know, I go once a week to go and have my supervisions and I try and coincide that with the department seminar. But now I kind of read groups and there's the randomised coffee trial, which is just genius. So actually, I probably got to meet more people after lockdown started, not because of just starting the postdoc. That's not really where the change happened. But just when everything went online, so.

Terry Soo  25:16  
So what do you plan to do? This postdoc goes on for...?

Clair Barnes  25:19  
Its another year and a half. So we finished September, I'm not sure which end of September September 2022.

Terry Soo  25:26  
Are you already applying for jobs? And if it's over in September, is it?

Clair Barnes  25:30  
Thesis first, then apply. Priorities.

Terry Soo  25:36  
Whatever, yeah.

Clair Barnes  25:37  
I'm gonna have a couple of months of not having to worry about a job. But yeah, I suspect I have to start applying for my next position alarmingly early, don't I? I'm kind of keeping an eye on job offers that kind of come through just getting an idea of what I might like to do, but I'm not really thinking about it actively.

Terry Soo  25:54  
You've gone to the interviews that we had here. Did you go to my interview?

Clair Barnes  25:59  
I don't know. I'm sure I would remember if I'd been in yours.

Terry Soo  26:02  
You could have you could attack me. You could have went to Richard. This guy. This guy's complete garbage. Don't hire this guy.

I'm sure he would remember I think because that was all happening when it was still in the department.

Yeah, yeah. It was June, June 2000. Like 19 or something like that.

Yeah, it was just before you started like the September before we all

started January. I started January.

Clair Barnes  26:26  
It was it was odd timing, I think. Yeah. So because it was all happening in the department. And I wasn't going into the department very much unless it happened to be on a day when I was going in because I had to do something else. I think I've sat in on like two days of the interview presentations. There was quite fun.

Terry Soo  26:42  
Yeah, because if this is what you in some sense, it's good experience to go to the interview. So you see what to expect, right? So people usually usually put on their A game when they go to their interviews, right? Otherwise, if it's just some regular talk, you don't you don't bother dressing up in the suit, but you bring your you bring your best suit. You bring your best suit to the interview usually right? Are you What do you guys call it here? I was interviewing somewhere. And they said that this could be a dinner at the bar. Okay, smart dress, something some smart dress not required for the dinner or something like this because the dinner was casual or something like this, but they use the word smart dress

Clair Barnes  27:18  
Wear your fancies garment. Did you have a good time doing a PhD here? Yeah, I really enjoyed it. I mean, I've stuck around for a postdoc. Can't have been that bad? Yeah, no, I think it was I did it a slightly unusual way and that I was working from home the whole time. So I haven't been as involved in the department is probably a lot of people would be, but actually, there was still quite a lot of opportunities to kind of meet people, I wouldn't say I felt super isolated, even only coming in kind of once a day, because we would always have the ETF, the department seminar, as I say, I try and arrange my supervisions to kind of coincide with that so that I could talk to other people, just once, once a week. And then you'd always have the department tea. So I got to know quite a few people in the department doing that, but just more staff than students. I think I didn't get to meet many of the PhD students. So yeah, I suspect if you actually came in and worked in the PhD room, which was something I didn't do, then then you'd have a lot more social interactions with with other people.

Terry Soo  28:16  
Do you have any advice for for future PhD students or just future statisticians in general?

Clair Barnes  28:24  
And my big one at the moment is finished your thesis before you start your postdoc? Thats my Golden Rule.

Terry Soo  28:30  
It was always just to get the next job. It doesn't matter what's the next that thing doesn't even matter. As far as I'm concerned, you get the next job. That's all that matters is the printer position. Right? That's all that matters.

Clair Barnes  28:41  
Permanent position will be different. Yeah, what else? I don't know. Just find a supervisor that you actually get on with, I guess, I get the impression a lot of students just find a project and kind of go Oh, that'll do. I'll just apply for that. Whereas I because I kind of did it in this slightly weirdly timed weird route by actually just getting in touch with Richard I kind of knew after I chatted to him that we had a similar sort of approach and similar sort of sense of humour. And we just kind of got on when I spoke to him. And I thought Yeah, I could probably deal with spending a few years working for you.

Terry Soo  29:15  
It's only three years, right? Or something like that, right? 

Clair Barnes  29:18  
But you really want it to be someone that you do actually get on with. So I would that would be my my advice.

Terry Soo  29:25  
What about other other young young statisticians in general? Do you have any sort of advice or any any any outlook words of wisdom? Your person is acquired the next job already right?

Clair Barnes  29:38  
Yeah, don't annoy your supervisor and maybe they when they see a postdoc opportunity they will recommend try not to avoid annoy your supervisor. That's probably the big one.

Terry Soo  29:48  
Well, thank you very much, Claire, for speaking to us tonight. And next time we speak hopefully, it'll be Dr. Barnes. And then soon enough after that, you know, in the UK, it's a big deal here right you guys have like this, this And this doctor Professor distinction right whereas every everyone in North America is Professor right and hopefully a few years after that it'd be Professor Barnes.

Clair Barnes  30:09  
That or Dame

Terry Soo  30:11  
Dame, Lord or whatever the Dame professor I don't know what order does it go does it go Professor Dame or Dame professor?

Oh, no. We can look this up needs to come to that one though before it becomes an issue. So I think we've got on that one.

All right. Well, thank you very much and good night. We'll speak to you again soon.

Clair Barnes  30:30  
Yes, Take care. Bye.

Unknown Speaker  30:35  
UCL minds brings together the knowledge, insights and ideas of our community through a wide range of events and activities that are open to everyone.

Transcribed by https://otter.ai