1 00:00:03,340 --> 00:00:08,410 OK, so welcome, Maria Kwacha and Maria, I'm going to hand over to you. 2 00:00:08,410 --> 00:00:17,470 And you can take it from there. We're going to hear all about the day in a day in the life of a statistics consultant. 3 00:00:17,470 --> 00:00:20,860 Yeah, yeah. Yeah. Big. 4 00:00:20,860 --> 00:00:28,150 Thank you very much. So guys, very mind that we are in the same room sitting at the same desk, so the logistic may be awkward. 5 00:00:28,150 --> 00:00:37,400 They're with us, please. OK, I'm going to show a scene. 6 00:00:37,400 --> 00:00:42,750 Yes. OK. And you all seem to say this full screen is everything, not excellent. 7 00:00:42,750 --> 00:00:47,330 Thank you, sir. Well, thank you so much for having us. I am Maria Christodoulou. 8 00:00:47,330 --> 00:00:53,910 This is my colleague, Megan. What's going on? We wanted to talk to you a little bit about things like this. 9 00:00:53,910 --> 00:01:02,660 Consulting on how we planning on doing this will give you a brief introduction of who we are in the consultancy and also where we are. 10 00:01:02,660 --> 00:01:08,130 For those who don't know what we do in general and who we work with, and then I, 11 00:01:08,130 --> 00:01:14,720 we will give you an overview of what our day to day existence looks like and some examples. 12 00:01:14,720 --> 00:01:20,570 And then we'll tell you the things we really love about itself because we're both quite enjoy our jobs. 13 00:01:20,570 --> 00:01:24,530 Then we're going to own up to the things we're not very good at. 14 00:01:24,530 --> 00:01:29,390 I'm tell you the things we really struggle with, and we're going to finish in telling you how to. 15 00:01:29,390 --> 00:01:36,120 You can get involved with the consultancy. Obviously, interrupt me if you have any questions, any comments, 16 00:01:36,120 --> 00:01:41,690 there's going to be a part that I'm hoping it's going to be interactive, so you're going to hate me for it. 17 00:01:41,690 --> 00:01:49,200 And those of you who are brave enough to pull your covers up on the two of you that says, I may pick on you, so. 18 00:01:49,200 --> 00:01:57,700 Excellent, so the. The consultancy is composed of three members at the moment, we have chromite Scott, who is the director of the consultancy, 19 00:01:57,700 --> 00:02:04,070 and she has extensive experience of almost two decades of working statistical consulting. 20 00:02:04,070 --> 00:02:13,360 She's calling you maternity leave, but we were waiting to see her back in June of leave, beginning of June. 21 00:02:13,360 --> 00:02:25,450 My colleague Mandy Ignacio Sosoli is a statistician and has quite a lot of expertise in all things that have to do with statistics, also statistics. 22 00:02:25,450 --> 00:02:35,740 He thought a lot of things over time and I am a little bit Paul, but I wear multiple hats some of time by myself to the other half of my time. 23 00:02:35,740 --> 00:02:41,880 I am a consultant and my expertise is mostly in biostatistics. 24 00:02:41,880 --> 00:02:50,880 So what do we do, what we do in general is we help on absolutely every aspect of our clients work. 25 00:02:50,880 --> 00:02:58,650 First of all, can I help the experiment the participants on so I can see somebody is giving me a blanket? 26 00:02:58,650 --> 00:03:04,930 Thank you. Thank you. So we can get involved in study design. 27 00:03:04,930 --> 00:03:10,770 So if somebody has a scientific question and the one to design an experiment, we can help them with the experimental design. 28 00:03:10,770 --> 00:03:14,730 We can give them advice on data collection. We obviously don't collect data for them, 29 00:03:14,730 --> 00:03:23,250 but we can guide them into doing it in a way that would give them the best chance of getting a robust statistical result. 30 00:03:23,250 --> 00:03:32,980 Often we have to step in with data management of data issues or reformatting the data so that it can be in a format that they can use for modelling. 31 00:03:32,980 --> 00:03:37,090 But the majority of our work is actually on data analysis, we handle data. 32 00:03:37,090 --> 00:03:44,350 We work with people's problems and we try to get answers to specific hypotheses that they have formed, 33 00:03:44,350 --> 00:03:52,680 and we write reports that they can then use in their in the publications for grant applications. 34 00:03:52,680 --> 00:03:59,750 All the things we do review advice and public insurance that's very prominent in the pro bono work we do, 35 00:03:59,750 --> 00:04:04,260 which many have not said we'll talk you through. And finally, 36 00:04:04,260 --> 00:04:09,210 but that's quite close to my heart because I know the majority of it we do training and 37 00:04:09,210 --> 00:04:14,700 that's statistics training for specific statisticians that need to use statistics. 38 00:04:14,700 --> 00:04:20,730 It's we understand that there is an issue between people using statistics poorly. 39 00:04:20,730 --> 00:04:25,230 But unless you go out of your way to train people to do this, rightly, 40 00:04:25,230 --> 00:04:32,070 there's very little that can be done to improve the situation, and we're working quite a lot on that. 41 00:04:32,070 --> 00:04:38,070 OK, I'm going to close the window again, guys, so you can see on the diagram right here. 42 00:04:38,070 --> 00:04:44,910 So basically what we take care of face the entire consultancy cycle as you can see, 43 00:04:44,910 --> 00:04:51,210 like we find clients and projects, so it can be that maybe we advertise ourselves, 44 00:04:51,210 --> 00:04:54,180 which is something that we are planning to do like in social media channel, 45 00:04:54,180 --> 00:05:00,330 for example, or that I had some clients recommend us to someone in their network. 46 00:05:00,330 --> 00:05:07,350 Or it can be someone from our personal network who get to us and asks for statistical support. 47 00:05:07,350 --> 00:05:15,750 Then what do we do? We try to really listen and and to clients formulate their problems because sometimes they do have a problem, 48 00:05:15,750 --> 00:05:22,950 but they don't know how to actually seek out, you know, what is the specific point in which they actually need help? 49 00:05:22,950 --> 00:05:26,700 Then often we have kind of done this requirement analysis. 50 00:05:26,700 --> 00:05:32,010 This escalated this way we design and agree and the scope of work. 51 00:05:32,010 --> 00:05:37,410 So what is the scope of work? Will detailed what quite a needle like. 52 00:05:37,410 --> 00:05:45,630 You know, it's just below the area of the city civil service and we are going to provide and we cannot like a few details about, for example, 53 00:05:45,630 --> 00:05:51,750 the modesty we are going to implement for them or if we're going to do some data cleaning and we can 54 00:05:51,750 --> 00:05:58,260 discuss about this with the client before we start the work and we wait for the client to sign off. 55 00:05:58,260 --> 00:06:02,910 Then we implement the work that we had sort of established in the scope of work. 56 00:06:02,910 --> 00:06:11,700 As Maria said, we can take care from the design up to the quality assurance of the results to the client had already. 57 00:06:11,700 --> 00:06:16,410 And so after we have implemented our work, we have done our analysis. 58 00:06:16,410 --> 00:06:18,000 We are prepared to report to them. 59 00:06:18,000 --> 00:06:26,360 We communicate findings to the clients just to make sure that we understand what we are the result of of our analyses. 60 00:06:26,360 --> 00:06:30,450 Sandy, we have done within depredation in this part of time. 61 00:06:30,450 --> 00:06:36,090 And then when we have the client sign off of the last deliverable added, 62 00:06:36,090 --> 00:06:46,080 we have provided them we do a sort of internal evaluation of how the project went to where we can kind of do better or we ask clients for feedback, 63 00:06:46,080 --> 00:06:51,420 actually. And it's something that we want to do in a sort of more systematic way. 64 00:06:51,420 --> 00:06:59,460 Yeah, we want to ask all the clients about this, about the feedback they can have regarding our work. 65 00:06:59,460 --> 00:07:07,650 And in this kind of cycle, there are a lot of activities that come on the side like we talked about on budgeting. 66 00:07:07,650 --> 00:07:13,710 So we need to take care of we need to understand which is that for us to adapt to certain project requires 67 00:07:13,710 --> 00:07:19,890 because we need to tell the client how many hours we will spend on that and how much it would cost for them. 68 00:07:19,890 --> 00:07:28,200 Then we take care of all the stakeholder relations, so we try to communicate and engage with the clients and also like with women 69 00:07:28,200 --> 00:07:32,520 internally to the Department of Statistics or another company we are working with, 70 00:07:32,520 --> 00:07:38,070 which is Oxford University innovation, which absence with the contextual side of our project. 71 00:07:38,070 --> 00:07:41,880 So we kind of keep the communication alive. 72 00:07:41,880 --> 00:07:48,390 And of course, it requires quite a good dose of team management and project management because we are not a big team, 73 00:07:48,390 --> 00:07:57,360 but also, like in all, trying to make two people work well together and make sure that everything is done timely. 74 00:07:57,360 --> 00:08:02,250 It's quite challenging. So why didn't we invest a lot of time in organising the work? 75 00:08:02,250 --> 00:08:06,180 And then, of course, we did have the communication that it's something we are still developing, 76 00:08:06,180 --> 00:08:10,800 as it was saying, because we don't really have a proper social media presence. 77 00:08:10,800 --> 00:08:17,990 But yeah, it's something that we are working on and hopefully in the near future, you will hear from us. 78 00:08:17,990 --> 00:08:27,680 So who are our clients? We kind of grouped them into main categories, we have internal clients meaning internal to the University of Oxford, 79 00:08:27,680 --> 00:08:33,740 and we have sort of classified ad hoc internal projects and pro-bono adult care represented 80 00:08:33,740 --> 00:08:39,560 by researchers or even field students who come to us with a very specific research question. 81 00:08:39,560 --> 00:08:48,290 And so we provide some help to them. Then we have like internal projects, and in this case, it's mainly research groups. 82 00:08:48,290 --> 00:08:53,750 They say across the university so they can come to us at the very beginning of their project, for example, 83 00:08:53,750 --> 00:08:59,900 when they are writing grant applications because they need, for example, help with the sample size computation. 84 00:08:59,900 --> 00:09:05,660 Or they can get back to us when you know they are more in their analytical part of the project. 85 00:09:05,660 --> 00:09:11,840 And so they need us to actually perform the analysis for them. Then we have the pro bono that Maria mentioned earlier, 86 00:09:11,840 --> 00:09:20,360 and it's what really makes us so proud of the work we do because we are nine clients at Administrative Department of the University, 87 00:09:20,360 --> 00:09:27,560 and we really try to ask them to improve their general level of statistics across the university. 88 00:09:27,560 --> 00:09:31,220 And, for example, a client we have like. 89 00:09:31,220 --> 00:09:40,250 They are exam companies. So everyone who is external from the University of Oxford, so companies are Universities International Organisation, 90 00:09:40,250 --> 00:09:47,930 and on this side, we work with all UI that I mentioned a few minutes ago, and they take care of the contractual side of our work. 91 00:09:47,930 --> 00:09:53,360 So they we ensure that everything is done properly, fairly transparently and so on. 92 00:09:53,360 --> 00:09:54,800 And then as you can see, we have the teaching, 93 00:09:54,800 --> 00:10:02,180 which is sort of just so because we can provide teaching both within the University of Oxford or to external organisation. 94 00:10:02,180 --> 00:10:07,160 And we have just a very recently like yesterday yesterday signed a contract with with 95 00:10:07,160 --> 00:10:13,580 an institution external to to the university to provide some specific training. 96 00:10:13,580 --> 00:10:15,590 So my day in the life, OK, 97 00:10:15,590 --> 00:10:24,660 we don't really have specific like a typical day because it very much depends on the projects you are working on our workload in good day. 98 00:10:24,660 --> 00:10:30,170 So we work on two or three production simultaneously in that day, so it can be even more. 99 00:10:30,170 --> 00:10:35,840 And of course, like each project, since we work, not quite they use of disciplines. 100 00:10:35,840 --> 00:10:40,250 Each project come with its own challenges. 101 00:10:40,250 --> 00:10:45,770 And so it requires a bit of time also to maybe get acquainted with the language sometimes. 102 00:10:45,770 --> 00:10:49,760 And then you can tackle this statistical problem at times. 103 00:10:49,760 --> 00:10:54,380 And of course, Maria and I would work very well because, you know, 104 00:10:54,380 --> 00:11:00,920 we we have different strengths that are useful for different stages of our project management cycle. 105 00:11:00,920 --> 00:11:08,690 And so, you know, we try to we have the content to make it so that you can tell the world that. 106 00:11:08,690 --> 00:11:16,580 So what happens when a query comes in on the first thing that happens, it's normally an email on some of these asking for specific question. 107 00:11:16,580 --> 00:11:24,140 We actually try to understand the general topic, and that would give us an idea of whether we have the vicinity of skills we need. 108 00:11:24,140 --> 00:11:28,400 Do we have the skills that we need or do we have absolutely no knowledge of it? 109 00:11:28,400 --> 00:11:32,900 And sometimes it can be a low skill from the first contact we don't really know. 110 00:11:32,900 --> 00:11:35,750 Other times, it's something very familiar. 111 00:11:35,750 --> 00:11:43,120 A few days ago received a query that these essentially what I did for my Ph.D., so we have more expertise in some things than other. 112 00:11:43,120 --> 00:11:53,300 Now, if we find that we don't have all the expertise we need, we still can work out whether we can get the necessary skills for the task. 113 00:11:53,300 --> 00:11:59,420 In a sense, it's a question of is this something that I can actually quickly learn to implement because I know 114 00:11:59,420 --> 00:12:05,290 sufficient statistics on my own to be able to understand the new method that's relevant to this. 115 00:12:05,290 --> 00:12:12,070 Then we try to get some clarification, professional timelines, for example, to very important, can we fit it into a workload? 116 00:12:12,070 --> 00:12:17,110 Our workload is quite varied in terms of what we have to do and when we have to do the things on there. 117 00:12:17,110 --> 00:12:22,570 Some weeks that are much more difficult to include new things in than others. 118 00:12:22,570 --> 00:12:29,950 But also we need to take into consideration when our clients will need delivery of the final output we're going to give them. 119 00:12:29,950 --> 00:12:35,620 Is this going to be too late or are we doing it at the right time for them? 120 00:12:35,620 --> 00:12:41,440 Normally we follow up by requesting a meeting with we have found that the timelines look OK. 121 00:12:41,440 --> 00:12:48,730 We we feel like we can provide some help for this. We actually request a meeting to get more in-depth understanding of the problem. 122 00:12:48,730 --> 00:12:57,790 On that meeting, we need to verify that we really are happy, we understand exactly what they need and that we're comfortable putting the query. 123 00:12:57,790 --> 00:13:04,150 In some cases, it may be something that's beyond our expertise, and we wouldn't be able to do a good job on it. 124 00:13:04,150 --> 00:13:13,240 So we need to run a quality assessment of what we will be able to provide, need to be very critical and harsh with ourselves. 125 00:13:13,240 --> 00:13:22,040 But we need to be able to ensure that everything we present is of the song that we want it to be, which is why time? 126 00:13:22,040 --> 00:13:31,430 One of thing is we it's about stage where we normally have a discussion in case of something that concerns us in terms of ethics, for example. 127 00:13:31,430 --> 00:13:35,150 That doesn't happen often, but some projects they want to. 128 00:13:35,150 --> 00:13:42,980 And that's often external projects, not in time, but they want to use findings in a particular way that makes us uncomfortable. 129 00:13:42,980 --> 00:13:51,080 And for that, we use our judgement and our own moral compasses and we make a decision on that. 130 00:13:51,080 --> 00:13:54,760 What we do have to that is we put together are all the money. 131 00:13:54,760 --> 00:13:58,400 You already said that we put our scope of work and that's exactly it. 132 00:13:58,400 --> 00:14:07,430 Essentially, we make a document that says these are the things that we're going to be doing to answer the question, but our clients want to answer. 133 00:14:07,430 --> 00:14:14,480 This is when we're going to be doing it and this is how much it's going to cost and we pass it on to our clients for an agreement. 134 00:14:14,480 --> 00:14:21,800 If it's an external company that then has to go through a UI for a formal contractual agreement. 135 00:14:21,800 --> 00:14:23,900 And this is what a query looks like. 136 00:14:23,900 --> 00:14:32,330 This is a fictional example, because this was the one that was given to me at my interview when I was interviewing my current job. 137 00:14:32,330 --> 00:14:37,850 This is what they look like. Yes, I think that's great. This is what they normally look like. 138 00:14:37,850 --> 00:14:39,700 It doesn't affect what they're saying. 139 00:14:39,700 --> 00:14:52,340 So it's like they come with a certain set of questions, and we need to quickly understand whether a this is something that we would probably help. 140 00:14:52,340 --> 00:14:59,900 And B, we need to decide what questions would we ask them in the follow up meeting to be able to really get 141 00:14:59,900 --> 00:15:05,270 enough information for us to make a scope of work and be able to carry on with our work for it? 142 00:15:05,270 --> 00:15:12,980 So this is the interactive part, I'm afraid, so feel free to either use the chat if you want in. 143 00:15:12,980 --> 00:15:19,540 Sorry, let's go plug. It's no free to use the chat and. 144 00:15:19,540 --> 00:15:28,870 And give up on that or unmute yourself. But what questions would you ask in a meeting if you went to see this particular query? 145 00:15:28,870 --> 00:15:37,990 What are the things that stand out to you? I want to be a consultant. 146 00:15:37,990 --> 00:15:47,350 Thank you, Garrett. We are being recorded. So that was a question I got very excited. 147 00:15:47,350 --> 00:15:57,800 Anybody brave enough, so quiet. Point one isn't exactly clear. 148 00:15:57,800 --> 00:16:15,370 Perfect, why isn't it what what makes it unclear? It it doesn't really read as English. 149 00:16:15,370 --> 00:16:22,590 But it's a bit out of context. Yeah, I was going to say, I feel I feel like my immediate thought is, 150 00:16:22,590 --> 00:16:27,640 can can we get a little bit more context about exactly what they're trying to achieve there? 151 00:16:27,640 --> 00:16:32,080 I mean, the thing you started talking when I was typing, the next thing I was going to type, 152 00:16:32,080 --> 00:16:36,610 which is the thing that's immediately jumping out here since you mentioned ethics previously, 153 00:16:36,610 --> 00:16:41,020 is they're talking about working with adolescents and school pupils. 154 00:16:41,020 --> 00:16:52,540 So in addition to sort of all of the usual ethics that goes with collecting people's data, you have you suddenly have enormous safeguarding issues. 155 00:16:52,540 --> 00:17:00,160 And so it's of additional ethical issues surrounding it. Are you are you going to be directly interacting with students? 156 00:17:00,160 --> 00:17:05,010 Are you going to be saying, you know, keep data on people under the age of 18? 157 00:17:05,010 --> 00:17:11,830 I'm sure there's a whole extra level of security that you have to go through there to make sure what you're doing is actually ethical. 158 00:17:11,830 --> 00:17:15,580 So above and beyond what you would normally do, just dealing with people's data. 159 00:17:15,580 --> 00:17:23,630 The topic is my sensitive is not just the fact that we have, in some cases, young adults or teenagers. 160 00:17:23,630 --> 00:17:30,580 It's the fact that we're looking at mental health, which could be it's this this could be vulnerable adults, 161 00:17:30,580 --> 00:17:34,240 but it's also protected characteristics that will be collected. 162 00:17:34,240 --> 00:17:41,920 So should this project be funded, we would have to see that they have a threat, for example, because that's an internal project. 163 00:17:41,920 --> 00:17:47,170 We have to see that they have passed an ethics bill to be able to do this and that we have been included. 164 00:17:47,170 --> 00:17:56,440 If we are to have a handle the data, we have been included and there have been specific, so specific. 165 00:17:56,440 --> 00:18:08,920 Up specific ways where we can access the data without compromising anybody's personal details, even if they are anonymous today. 166 00:18:08,920 --> 00:18:14,400 We still need to protect them, respect them. The other thing is that. 167 00:18:14,400 --> 00:18:21,030 It's very difficult to actually understand what kind of analysis this would mean. 168 00:18:21,030 --> 00:18:27,930 This is quite close to your expertise in terms of models, so when you saw a query like that, 169 00:18:27,930 --> 00:18:31,770 what would be the actual statistical questions you would ask yourself? 170 00:18:31,770 --> 00:18:36,990 Well, first of all, I think that's what Fergus was saying. It's really like it's a bit out of context. 171 00:18:36,990 --> 00:18:43,500 So one of the first questions they would ask would be like, Can can I have more information about your project? 172 00:18:43,500 --> 00:18:45,270 And in general, when you let them speak, 173 00:18:45,270 --> 00:18:53,040 you can kind of pick up elements that you know are out for you to understand which kind of statistical model then you can, 174 00:18:53,040 --> 00:19:00,120 which kind of statistical problem they may have in order to answer to this specific question. 175 00:19:00,120 --> 00:19:07,700 And in particular, what they find useful sometimes is actually, I kind of process what they're telling me, what they're saying to me now. 176 00:19:07,700 --> 00:19:13,830 So you meant that, for example, wanted to you have a radius scale, which is measuring these properties for adults. 177 00:19:13,830 --> 00:19:20,590 So you want to be that's what adolescents do. You have any other, for example, example when they're doing what they're doing that? 178 00:19:20,590 --> 00:19:29,070 Can you find an example, you teacher, that you should have the same measures can be used for both adults and children when I got this sense. 179 00:19:29,070 --> 00:19:34,620 And yeah, in this case is a very specific question because it's up psychometric scale. 180 00:19:34,620 --> 00:19:38,370 So it requires very specific kind of analysis that need to be done. 181 00:19:38,370 --> 00:19:46,350 So you want, for example, one thing that you want to measure is are the items in question of actually measuring what I want to measure. 182 00:19:46,350 --> 00:19:52,770 So you have like a sort of latent construct that is correlating with these items and you need to kind of understand this. 183 00:19:52,770 --> 00:19:56,550 It's exactly what thereafter is exactly what it is doing. 184 00:19:56,550 --> 00:20:02,190 And then of course, there are a lot of other kind of you can go more or less enough see depends on, 185 00:20:02,190 --> 00:20:07,740 let's say, the first meeting with the client that you're going to have after having received it. 186 00:20:07,740 --> 00:20:13,110 But this is just to give you an idea of which the kind of queries we receive this is the thing. 187 00:20:13,110 --> 00:20:21,810 The thing with these things is that quite often for interactions, the majority of what we need to do is we need to use soft skills to get answers to. 188 00:20:21,810 --> 00:20:29,640 Statistical questions were not been asked directly because if I spoke to somebody in a sciences and I start 189 00:20:29,640 --> 00:20:35,820 asking questions about effect size where there may not have considered it does affect size in that sense. 190 00:20:35,820 --> 00:20:42,780 I may get a confusing answer if I start asking them, Is anyone else found this on what were the sort of numbers that we're getting from them? 191 00:20:42,780 --> 00:20:51,000 So you need to often asking direct questions to get very clear answers because you don't also don't want to be the main statistician on the call. 192 00:20:51,000 --> 00:20:58,680 That's terrifying everyone. You want them to relax and tell you about the project and force you with the actual results. 193 00:20:58,680 --> 00:21:02,850 But this is this is what it looks like. Authoritative time on. This is what we try to do. 194 00:21:02,850 --> 00:21:11,610 We look at this and say the first thing I would say is, OK, this needs we need to see if that has a wreck, as is how the fix is responded. 195 00:21:11,610 --> 00:21:15,900 What is the actual context of this project? And then we're thinking this is psychometric scales. 196 00:21:15,900 --> 00:21:18,270 We need to know whether they have, 197 00:21:18,270 --> 00:21:24,330 whether the property is basically that they're trying to actually estimate and whether they have been properly calibrated in that sense. 198 00:21:24,330 --> 00:21:35,850 Let's say that the first reaction is like. Well, what is the then we have what kind of reaction that hype? 199 00:21:35,850 --> 00:21:39,470 So what do we do like how do we deal with that? 200 00:21:39,470 --> 00:21:45,500 We received questions like email. We say, what? Then we try to read it and understand what can be the questions. 201 00:21:45,500 --> 00:21:50,510 And then we want to ask the clients because we do have quite a bit of preparation before that. 202 00:21:50,510 --> 00:21:55,010 And in general, we kind of we really discuss a lot between each other. 203 00:21:55,010 --> 00:22:00,810 We because maybe one of us can have can, not just something that the other person noticed. 204 00:22:00,810 --> 00:22:07,190 And I think it's really valuable. And then as soon as we have agreed with the client clarified, 205 00:22:07,190 --> 00:22:13,310 all our initial questions define what we are going to do to actually answer to their problem to tackle this problem. 206 00:22:13,310 --> 00:22:20,420 Then we start working on that and we make a point of updating regularly to client and what we are doing because, 207 00:22:20,420 --> 00:22:26,360 you know, sometimes you make estimates of what is the time that you need to perform to perform a task, 208 00:22:26,360 --> 00:22:31,520 but then you realise that the data are messy, that you would expect change and the model is quite small, 209 00:22:31,520 --> 00:22:37,160 is more challenging than you would expecting because there is some something little that really doesn't work. 210 00:22:37,160 --> 00:22:42,800 And so, you know, you kind of keep on, keep them posted and, you know, so they know where you are. 211 00:22:42,800 --> 00:22:48,590 They know if you need anything from them and you know how to count how you can move forward. 212 00:22:48,590 --> 00:22:54,170 And in general, our output is it's a report that we can write to, 213 00:22:54,170 --> 00:23:00,380 even like it can be very extensive or it can be just bullet points providing some insights and insights. 214 00:23:00,380 --> 00:23:04,700 And we also provide the code, which in general is almost on file, 215 00:23:04,700 --> 00:23:12,140 which we use to generate the report so they can find the code with what they need when they need it, and they can rerun the analysis. 216 00:23:12,140 --> 00:23:20,540 We think it's a good practise because some it gives us a spot in Sichuan for work and they can go ask questions about that. 217 00:23:20,540 --> 00:23:28,640 And it's also a way for providing a sort of training on specific maybe functions and comments or not. 218 00:23:28,640 --> 00:23:32,670 So we wanted to give you just an overview of other few cases that we have. 219 00:23:32,670 --> 00:23:38,390 We have worked on and we have divided them by kind of clients that we have, for example, the general case studies. 220 00:23:38,390 --> 00:23:43,580 They come from two departments in the university. The first one is computational biology. 221 00:23:43,580 --> 00:23:45,710 The other one is sociology. 222 00:23:45,710 --> 00:23:53,150 In the first one that is actually came to us with a very specific research question, which was understanding the role of geography, 223 00:23:53,150 --> 00:24:01,970 so the geographical location and the genetic data of an individual in their survival mechanism and in particular, 224 00:24:01,970 --> 00:24:09,920 they were interested in how you can yield meaning all the elements belonging is coming from a population such as Neanderthals, 225 00:24:09,920 --> 00:24:20,420 Denisovans, and they came to us with our with their database, which we have a few concerns about the quality level. 226 00:24:20,420 --> 00:24:28,190 And so we we have to work around the data and try to find out which was the best subset to actually 227 00:24:28,190 --> 00:24:33,970 respond to this question and be confident on the quality of the estimate we would have got. 228 00:24:33,970 --> 00:24:42,990 And now, after quite a lot of working and communication and changes, we are in the stage of writing a paper with them. 229 00:24:42,990 --> 00:24:52,850 So it was really rewarding as a project. And then the other one was an experiment from its a peculiar design of this experiment, 230 00:24:52,850 --> 00:24:59,030 and they needed someone to ask them to find the optimal number of scenarios they had to 231 00:24:59,030 --> 00:25:05,120 present to their respondents so they could have robust results out of their survey. 232 00:25:05,120 --> 00:25:08,300 So this one, this is quite challenging. We're still working on that. 233 00:25:08,300 --> 00:25:13,760 But you know, we're kind of progressing and this was an opportunity for us to learn a new technique. 234 00:25:13,760 --> 00:25:18,570 So the client, like we did, have discussions in the beginning. We were very upfront with the client. 235 00:25:18,570 --> 00:25:25,850 So we really value honesty, like saying, Look, I don't have the skills, but if you can provide training, we can absolutely help. 236 00:25:25,850 --> 00:25:33,290 So they had someone in the research group who was working on that and actually at the moment was too busy to talk to us to keep on asking them. 237 00:25:33,290 --> 00:25:41,330 So they provided training to us and we had the chance to learn something new, offered them a reduced fee, of course. 238 00:25:41,330 --> 00:25:47,690 And then, you know, I'll give them the results they were waiting for. Then we had some similar case studies. 239 00:25:47,690 --> 00:25:53,750 The first one is the university, the second one, these are medical. These are medical company based in Oxfordshire. 240 00:25:53,750 --> 00:26:00,860 And in particular, the first one was quite interesting because the researcher came to us, but they didn't have a very specific research question. 241 00:26:00,860 --> 00:26:03,590 So we asked them to looking at the data, 242 00:26:03,590 --> 00:26:12,290 try to fine tune their research question they could have wanted to answer to or they may have found interesting. 243 00:26:12,290 --> 00:26:19,970 And in particular, it was a very interesting project because I was about them analysing the genetics, 244 00:26:19,970 --> 00:26:25,490 the generating process of errors in computer systems, assisted translation studies. 245 00:26:25,490 --> 00:26:30,030 So it's not like. Something that you can find every day to work with, 246 00:26:30,030 --> 00:26:35,430 and the other case was they had this medical company had created a new biomarker 247 00:26:35,430 --> 00:26:42,060 and they use it just to diagnose together with some recognised biomarkers too. 248 00:26:42,060 --> 00:26:49,350 They use it to diagnose certain conditions, and they wanted to see if there were any interesting options in their data. 249 00:26:49,350 --> 00:26:54,720 So we have done with the analysis and received quite a good feedback from them, actually. 250 00:26:54,720 --> 00:27:02,100 Then we have done teaching online that any person we like, as you can see outside this case is quite fabulous. 251 00:27:02,100 --> 00:27:06,780 So one was the Department of Psychiatry. The other one was a CDC data science. 252 00:27:06,780 --> 00:27:11,410 I was four with the depicted institute and it was really, really interesting. 253 00:27:11,410 --> 00:27:18,600 So we had a different kind of audience because the first one was with people who didn't have a probability, much training. 254 00:27:18,600 --> 00:27:24,760 So they didn't have a very sound, much background while the other one was with people who had a solid background. 255 00:27:24,760 --> 00:27:28,380 Then it was sort of funny trying to tweak the content and, you know, 256 00:27:28,380 --> 00:27:33,420 just select the audience and then they put a bomb on one just to give you an idea. 257 00:27:33,420 --> 00:27:36,390 The first one is from the financial department of the university. 258 00:27:36,390 --> 00:27:43,080 The academic activity surveys a very important survey, which comes in every every year, 259 00:27:43,080 --> 00:27:48,540 and it's very important because the university asked to communicate the results to the UK government. 260 00:27:48,540 --> 00:27:56,410 So they wanted to have an idea about the robustness and the quality of disability they were going to deliver. 261 00:27:56,410 --> 00:28:05,070 I just to their to the academic staff. And we didn't find any kind of concern from us to the way in which this review was structured, 262 00:28:05,070 --> 00:28:10,550 but we realised they didn't have a specific kind of they didn't. 263 00:28:10,550 --> 00:28:17,940 They didn't have accommodation for some of the disabled staff, but they may be actually quite savvy. 264 00:28:17,940 --> 00:28:27,180 We had sort of equality and diversity concerns, so we faced with the equality and diversity and inclusion of what we see of the university. 265 00:28:27,180 --> 00:28:35,340 And then we worked with the alumni database, which and it brings together all the data from the alumni, from different colleges. 266 00:28:35,340 --> 00:28:41,190 So you can imagine the quality of this data and the person who was working with that was really concerned. 267 00:28:41,190 --> 00:28:47,670 So it again to us because of course, as you might imagine, they used this database of for funding reasons. 268 00:28:47,670 --> 00:28:58,050 So it is a very important role, and it's really important that the quality of the data in the database is quite high. 269 00:28:58,050 --> 00:29:03,030 So let's talk about the things we really love about our jobs, and there are quite a few of them. 270 00:29:03,030 --> 00:29:08,970 But for both of us, the most important thing is it's so diverse, it's constantly changing. 271 00:29:08,970 --> 00:29:14,070 We're learning new things all the time. And no two days with all the same. 272 00:29:14,070 --> 00:29:21,720 We haven't had a project that is identical to something else. We'll just take the solution we had for someone on reapplied. 273 00:29:21,720 --> 00:29:27,030 Every day is said to be something new and we are learning so much through this. 274 00:29:27,030 --> 00:29:38,580 We also become better communicators, especially since we have to communicate quite some quite nuanced ideas and concepts, 275 00:29:38,580 --> 00:29:45,180 but we have to do it in a way that is correct for the actual discipline. 276 00:29:45,180 --> 00:29:52,470 For example, my agnostic you mentioned earlier that the first bit we may even have to learn the language of the discipline. 277 00:29:52,470 --> 00:29:57,030 Multiple disciplines could be using a very similar technique, but they're calling it different ways. 278 00:29:57,030 --> 00:30:01,950 Or they may have some standards that they follow, not just on paper, in any other discipline. 279 00:30:01,950 --> 00:30:11,370 We have to limit them to be able to communicate with the researchers and understand what it is they want and whether we can provide it appropriately. 280 00:30:11,370 --> 00:30:16,240 We also feel we make a difference in the university every time we have an ad hoc 281 00:30:16,240 --> 00:30:21,670 query and there is a full student somewhere who is not sobbing next to the data. 282 00:30:21,670 --> 00:30:30,340 That's a win. Every time we have a pro bono where we actually help make this, make this universe become more solid statistics, 283 00:30:30,340 --> 00:30:35,040 that's a win for us every time we teach somebody to do something they didn't know how to do. 284 00:30:35,040 --> 00:30:41,190 Again, a win. So we make a difference and we do meet wonderful people and we meet some exceptional researchers. 285 00:30:41,190 --> 00:30:45,580 I it is really a joy. I'll find talking to joy. 286 00:30:45,580 --> 00:30:54,420 And there are some things we don't love, and we don't love them because some of them are our fault and we admit to that. 287 00:30:54,420 --> 00:30:58,840 Others, they just can't be helped. We do have the usual suspects. 288 00:30:58,840 --> 00:31:06,360 Our workload is quite heavy and. As careful and those horses as we are with our scheduling, 289 00:31:06,360 --> 00:31:13,350 we are at some point going to end up having the same deadline within two deadlines within the same hour, 290 00:31:13,350 --> 00:31:16,830 and we're going to be panicking and there will be tears and that does happen. 291 00:31:16,830 --> 00:31:21,310 It doesn't happen often, but it does happen. We also have competing demands, 292 00:31:21,310 --> 00:31:28,220 sometimes we need to decide to put the project aside when we're making good progress in it because something else is more urgent. 293 00:31:28,220 --> 00:31:31,070 And then when we pick up the project we put aside, 294 00:31:31,070 --> 00:31:37,330 we have forgotten half of what we're doing and it needs extra time and it's a little bit frustrating. 295 00:31:37,330 --> 00:31:46,190 One of the problems we have on that is down to us to get better if it is accurately estimating how long things will take, 296 00:31:46,190 --> 00:31:50,900 those takes longer than we think, even when we think. 297 00:31:50,900 --> 00:31:59,450 Who are the buffer we go over the buffer sometimes. That's because we don't often know what we're going to find until we get in there, 298 00:31:59,450 --> 00:32:08,550 and we have already agreed contracts in all terms of reference, how long we think it's going to take before we have the chance to really. 299 00:32:08,550 --> 00:32:12,420 Get properly inside the programme and try the modelling we have. 300 00:32:12,420 --> 00:32:20,760 The only time we know how long it's going to take after Italy is after we've invested pretty much half the time we estimated for it. 301 00:32:20,760 --> 00:32:26,610 We also try to do too much in limited time, that's probably personality based, 302 00:32:26,610 --> 00:32:33,740 we have both have tendencies for perfectionism, which means we also overthink and over check. 303 00:32:33,740 --> 00:32:40,080 You mentioned the project of the values and that is a really wonderful project. 304 00:32:40,080 --> 00:32:43,860 What they needed to do is they wanted to do a canonical Barrett's analysis. 305 00:32:43,860 --> 00:32:52,800 That's analysis as everyone agrees that it should have one assumption of the East and that the assumption is multivariate normality. 306 00:32:52,800 --> 00:32:58,530 So the data were all relatively low quality and they were not behaving at all. 307 00:32:58,530 --> 00:33:04,050 So we start transforming and changing things around and trying to fix it. 308 00:33:04,050 --> 00:33:09,870 But we ended up checking and doing different types of motivated normal tests, 309 00:33:09,870 --> 00:33:15,900 and we ended up doing six different types of multivariate normality test because one just wouldn't do for some reason. 310 00:33:15,900 --> 00:33:22,830 So what we need to learn is we need to learn how to control our tendency to overthink on over tech. 311 00:33:22,830 --> 00:33:26,430 Because let's be honest, when we did six test, 312 00:33:26,430 --> 00:33:32,100 we didn't get more information than if we had done just the one that told us there is nothing good here, 313 00:33:32,100 --> 00:33:38,400 so we need to learn to be better with our own tendency to overdo it. 314 00:33:38,400 --> 00:33:43,320 And the good thing on this is that you can actually kind of improve by time to time. 315 00:33:43,320 --> 00:33:48,960 So, you know, now I feel much better on that. Did I used to feel, let's say, six months ago? 316 00:33:48,960 --> 00:33:52,950 So yeah, it's a kind of steep learning curve, but it's exciting. 317 00:33:52,950 --> 00:34:02,890 It's up. We have a bit of a buddy system that we try to set up where we tried to check each other's instincts when they're not particularly helpful. 318 00:34:02,890 --> 00:34:10,960 So the final part of our talk is how you could actually get involved if you want to and now will just give you a brief overview, 319 00:34:10,960 --> 00:34:14,260 and we're going to talk more specifically about things you can do, 320 00:34:14,260 --> 00:34:21,880 but you can get involved with any part of our work that you're interested in teaching, including concerning actual consultancy. 321 00:34:21,880 --> 00:34:23,740 You can get involved in all of this and you can be. 322 00:34:23,740 --> 00:34:33,190 The sorting puts off a couple of hours or maybe two a day, or you could actually provide consultancy service as part of what we're doing, 323 00:34:33,190 --> 00:34:37,210 which could be more than just a couple of hours you can teach with us. 324 00:34:37,210 --> 00:34:42,670 You can call mine either be this will teach a specific field if you want to get an experience or 325 00:34:42,670 --> 00:34:47,920 teaching or communicating with audiences that you may not be so familiar when teaching for the mean, 326 00:34:47,920 --> 00:34:54,480 it's a different set of skills to meet. And you can help us by collaborating on satellite projects, 327 00:34:54,480 --> 00:35:03,780 and that can be very intensive level of effort, it can be from a couple of days to quite a few days. 328 00:35:03,780 --> 00:35:05,700 Yes, basically, 329 00:35:05,700 --> 00:35:14,490 I wanted to add very quickly on what Maria said on the other slides that we have a system in place at the department to currently do so when 330 00:35:14,490 --> 00:35:24,570 we receive a call that we may think like we think may be of interest for any of the details or annual academic stuff in the department. 331 00:35:24,570 --> 00:35:29,310 We send a sort of call out to everyone, so you should all receive it. 332 00:35:29,310 --> 00:35:37,080 We have not been sending many lately, but you know, we found that you should receive it and then it's open to anyone and the person, is there. 333 00:35:37,080 --> 00:35:44,670 Someone who is interested can sort of contact us by email and then we have something put in place with the HRA. 334 00:35:44,670 --> 00:35:50,130 So there is a proper sort of contract that has been signed between the consultancy and to give you. 335 00:35:50,130 --> 00:35:55,230 And of course, it gives the chance to their student to earn some money. 336 00:35:55,230 --> 00:35:59,220 And yes, for the for the time, you are certainly investing in that. 337 00:35:59,220 --> 00:36:07,710 So it's something really to keep an eye out. And, you know, just see if maybe something that, as Maria was saying, is exactly your Ph.D. 338 00:36:07,710 --> 00:36:12,060 So maybe just like minimum maths for maximum gain. 339 00:36:12,060 --> 00:36:18,120 And all we can say that you are interested in acting out like you can send an email to us 340 00:36:18,120 --> 00:36:23,550 personally or we have to statistical consultants email or you can come in the room 310. 341 00:36:23,550 --> 00:36:28,080 We have a lot of good instances which can be so chocolate. Yeah. 342 00:36:28,080 --> 00:36:31,480 And yeah, you can just let us know what your radius of expertise are. 343 00:36:31,480 --> 00:36:37,500 So, you know, maybe we can we can kind of see if there's something that may be of interest for you. 344 00:36:37,500 --> 00:36:41,700 Then we have some future plans that we will like to ask about you. 345 00:36:41,700 --> 00:36:44,100 Like just have your opinion on that. 346 00:36:44,100 --> 00:36:51,420 So we will have to provide some consultancy training in the department, maybe something like more extensive than the stock, 347 00:36:51,420 --> 00:36:57,120 like a week in the life of a statistical consultant where we can sort of talk of the consultancy cycle, 348 00:36:57,120 --> 00:37:04,750 explore every aspect and select do something together, like if we see a query from a client or we, 349 00:37:04,750 --> 00:37:09,000 we can create a mock study and then we can work together on each stage. 350 00:37:09,000 --> 00:37:14,010 And just as you gain these skills. 351 00:37:14,010 --> 00:37:18,540 Another thing that we wanted to do, which is more like we are therefore supporting you back, 352 00:37:18,540 --> 00:37:25,230 is more you are on your own so we can leave you in the wild, which is the statistical consultancy challenge. 353 00:37:25,230 --> 00:37:33,690 So we would like to create like groups of feet for people and give to each Real-World statistical query like Consultancy Way. 354 00:37:33,690 --> 00:37:36,210 And then, of course, you will have support from us. 355 00:37:36,210 --> 00:37:45,030 So we would like to find you on the scope of work like, you know, try the questions that you want to ask to the client and to with analysis. 356 00:37:45,030 --> 00:37:48,090 Maybe we always work on a peer review basis. 357 00:37:48,090 --> 00:37:54,990 And it's always helpful so we can provide this for you and we can be there when you are having the meeting with the client. 358 00:37:54,990 --> 00:38:02,310 And, you know, just ask you throughout the process and something that we wanted to do as well. 359 00:38:02,310 --> 00:38:07,830 And this is sort of building blocks, lunchtime data, carpentry workshops. 360 00:38:07,830 --> 00:38:14,490 So we would like to actually give everyone the skills to be able to. 361 00:38:14,490 --> 00:38:15,920 Transform that data. 362 00:38:15,920 --> 00:38:25,260 And like in the way that they would need to, and maybe they don't know to do that at the moment, so we thought they could be something an addition. 363 00:38:25,260 --> 00:38:32,520 Maybe someone already knows that we could be still interesting swing by seeing if there's anything else they can learn or they can. 364 00:38:32,520 --> 00:38:37,530 There's something new. And why would you want to get involved? 365 00:38:37,530 --> 00:38:47,250 That's the question. Well, it's exciting. I mean, if you have not ever at to from there, the presentation is very much our fault. 366 00:38:47,250 --> 00:38:53,040 But, you know, it's really exciting and you that learning curve is really steep, 367 00:38:53,040 --> 00:38:57,930 but it's really rewarding because you actually learn something new almost every day. 368 00:38:57,930 --> 00:39:07,050 And it's something that I personally really value like. Like, I'm building my toolkit and that it can apply to two different disciplines. 369 00:39:07,050 --> 00:39:13,350 And it's really it's really interesting. I feel I've become a better researcher since starting from the consultancy because 370 00:39:13,350 --> 00:39:17,100 I'm gaining knowledge in things that are actually relevant to my research. 371 00:39:17,100 --> 00:39:21,990 I just wouldn't have gone so far out of my actual research area to find them. 372 00:39:21,990 --> 00:39:25,170 So it actually helps with research as well. 373 00:39:25,170 --> 00:39:33,840 And the I, as we were saying, the work experience so you can have a little income, which is that I was at the moment for intimate consultancies, 374 00:39:33,840 --> 00:39:39,450 but externally, defensively, much of the client and what they need them for how long they would need someone. 375 00:39:39,450 --> 00:39:46,920 And of course, another thing is that you develop a lot of soft skills because you collaborate with others so you don't to communicate with clients. 376 00:39:46,920 --> 00:39:55,320 You learn how to communicate with your team members and just you make sure that everything runs smoothly and you do a lot of work on yourself as well. 377 00:39:55,320 --> 00:40:00,160 As Maria was saying, I was saying earlier, you know, how to really say to like, 378 00:40:00,160 --> 00:40:09,480 make sure that you don't overthink over check things and you are just like you get better every day defying deadlines, for example. 379 00:40:09,480 --> 00:40:15,030 And yeah, that's all from us. So thank you very much for listening. 380 00:40:15,030 --> 00:40:19,136 I was up Seamus Green, OK?