Sunday, August 14, 2016

Top Tips for PhD students

William Buchanan

Professor at Napier University


I have done many MSc/PhD examinations, and I enjoy them so much. With PhD thesis' I especially like seeing the face of the candidate when I take the thesis out of my bag, and you can't see the thesis for Post-it notes. This should hopefully show that I've read single word in it, and, hopefully, understood most of it. 
So here's my top Ten 25 hints on how to help your examiner (many of these should also be relevant to MSc thesis' too - to a student aiming at a 1st class Hons dissertation):
  1. Say up-front what the problem is, what other people have done, and how you have added to it. The Introduction chapter is the most important chapter of all, and you need to grab the reader, and tell them what the problem is, and how you have solved it. If the examiner understands the thesis after the Introduction, you are half way there. I've read a few thesis' which I had no idea what the point was until I actually got to the end, and the contribution was finally revealed on the last page. This is a major gamble, as some readers may give up before that point, and not know the end contribution. Be fair on the reader and tell them the contribution, and keep telling them.
  2. Get rid of those typos! You wouldn't believe the number of PhD thesis' I have read that have a typo in the very first line of the thesis. A reader becomes annoyed if they have to keep correcting  typos, and the more annoyed the reader, the more time they are taking away from actually reading the content. Try and start off on a good footing, so that the Abstract and Introduction chapter have been read over several times - typically talking it out loud. If possible get someone else to read the Introduction, and see if they understand what the point of work is.
  3. Bad grammar shows bad practice and a weak supervision. Part of doing a PhD is learning how to write and present ideas, and how to review and edit. One of the most important things that you learn in a PhD is how to write, so that others can understand your ideas. A good part of this is for supervisors to get involved reading the work, and in giving detailed feedback. It is often a good idea for supervisors to mark up early drafts with red pen, so that students get an idea on the amount of checking and editing that is often required.
  4. Superlatives are not very good! A PhD is a scientific study, and the usage of superlatives should be avoided, along with weak words like "big" ... "the measure gives a very big number". If a number is large, then define what large actually means, as everything is relative.
  5. Significance matters. I've read thesis' that draws a graph, and then gives me values of 10 decimal places, and then to be told that there is an importance of one thing to another. But is it significant? If I move from 100.01254632 to 100.1263241, is that a massive change and why do we need so many decimal places? Every measurement has an accuracy, and this should always be included in the presentation of the figures. Examiners want to know the significance of something, and if it isn't significant, then tell them.
  6. One table tells much more than a whole lot of numbers. Again I've read so many thesis', where the writer continually presents a whole series of numbers and graphs, and where they could all be moved into an appendix, and compiled into a single table (or graph). A good supervisor should be able to spot how to collapse lots series of tables into a single one, as they often have to do it for papers. Many students rely on drawing graphics for presenting trends whereas tables are often better, especially in defining changes within the figures presented. A great tip is to normalise values, and show how the values vary between each other. Relative values are often easier to understand than absolute ones - remember too that most values have units, and that units matter. I've quizzed many students on whether they are talking about Mbps or MBps - there's a difference of eight in there!
  7. Draw some pictures. There is no place for trivial graphics and clip art in a PhD thesis, but there is a place for the abstraction of complex ideas, especially in the introduction. There no real need to just copy the graphics from others, as they should come from the ideas inspired by the writer. I've read quite a few thesis, where the text just goes on and on. Break the text up every now and then, and give the reader something to ponder over. 
  8. Break up and but keep a narrative. There's a careful balance here. If you keep your sections short, it becomes to "bitty", and if you make them too long, they become long and unwieldy.  I personally read whole sections in a single sitting, and try and take in the ideas, and I won't move on until I understand it. A long section, especially where there are no sub-sections, often introduces too many concepts which can make it difficult to read.  I normally recommend a maximum of a page and a half of text before there should be a break (such as a sub-section break). Long paragraphs are not a good thing as it becomes difficult to take in all the concepts introduced. Try, if possible, not to make them too short, but not too long. A paragraph that goes on for half a page is probably too long, and one that has only two sentences is probably too short. Along try and avoid too many sub-sub sections, as it becomes difficult for the reader to put it all into context.
  9. Avoid using the words of others too much. A thesis is written by the writer, and it is their words. Long series of break-out text items of quoted material becomes fairly generic, where you get little of the sense of the thoughts of the writer. If you must reference others, pull it out, and indent.
  10. Be precise. A PhD thesis should be a scientific document which abides to certain standards for the articulation of ideas. It is always sloppy to see a candidate writing 9*6^3, where the "*" is a sloppy way of writing a multiply symbol (x) and ^ should be "to the power off". If it's an equation, it should be pulled out of the text, and a proper equation editor should be used, and a proper numbering system.
  11. Every diagram and table should be referenced in the text. I have read some many thesis (typically drafts) where the writer just assumes that the reader know how a diagram or table should integrate with the narrative. Every figure and table should thus be referenced in the text, so that the reader knows when to look at it. If possible don't break up your narrative with a diagram, and move it a little later on, as long as it is after then text which is referring to it.  
  12. Be critical of yourself and others. One of the key things within a PhD is the ability to critically appraise the work of others, both for the strengths and weaknesses of their work, and also of the candidate's work. I often circle the first signs of critical thought in a thesis. If it happens on Page 50, there's a problem in not being able to critically appraise work. Along with this some candidates can think that everything is perfect with their work, and that it addresses every single problem in their field. Try to always define both the strengths and weaknesses your own work, and identify how these could be improved. The scope of the impact should never be overestimated, but also not underestimated. If you've developed something that completely changed something, be up-front and tell the reader. Most of the times, be honest to say that you are just enhancing something a little bit.
  13. A thesis is not a diary! I have read so many thesis' which are basically just a chronological flow of their research. You can often spot this as the literature review runs out of references which are up-to-date. I have read several thesis' where the latest reference in the literature review is two years ago, and it points to the fact that it has not been updated since it was initially written. A literature review should be written for the thesis, and many parts of the original literature can be dumped, and replaced with newer references which fit in with the contribution.
  14. Focus the literature review on the contribution. One literature review of PhD thesis I read was almost 200 pages long, and my head was spinning at the end of it. It covered so many points, and few of them actually went anywhere in the following chapters. Try and focus the literature review on covering the 4 or 5 key concepts involved in the thesis, and not in the research project. A good supervisor can often spot  redundant sections, and advise for them to be cut. If the thesis is still the same by taking something out, there's no need for it to be there, as every paragraph and every word should count, and be carefully crafted as part of the whole story.
  15. Make sure the aim is of the thesis, and not of the initial research project.  Many thesis' start with "The aim of this research project is ..." which often is a sign that the original project aim has not changed in the writing of the thesis. Overall the aim is the aim of the thesis, as the research project has finished. Every should be written from a point-of-view that the work has finished, and this is the write-up.
  16. Get the flow right. A strong flow of literature, method, build and evaluation helps the flow of the thesis, and where you often see references to literature tailing off as the thesis develops. I've seen some thesis' where there are whole chapters that lack any form of reference to other work. This is poor practice as a PhD thesis should show how every aspect of the work fits in with the work of others. I like to see a reference to other work in the introduction of a chapter, as it shows some key influences for the work. I personally don't like an introduction that says "Section 1 says this, and Section 2 says that, and Section 3 says something else", as I can see from the table of contents what the contents are. If possible the reader should tell the reader what is likely to be revealed and what the significance is. A re-enforcement of the main drive of the work also helps to bring the focus onto the main contribution of the work.
  17. If you don't know it ... don't say it! This one seems so obvious, but you won't believe the number of times that you ask in a Viva about the detail of a paper, and the method used, and for the answer to be that they don't actually know what it does. You always increase your exposure to probing if you include things you don't quite understand, so dump them (if they are not a core topic). 
  18. Explain it simply. There's nothing nicer for an examiner when the candidate takes a complex idea and gives their own viewpoint on it, in a simple way, using new material. It shows that they can articulate complex ideas in a simple way. The standard test for any thesis is that a 14-year old child should, at least, be able to read it, and understand some of the key concepts in it.
  19. Show that you love the subject and that it is relevant.  Three years is a long time, but the sustainment of interest is a key part of the work, so try and show that this is an important topic and that your thesis is exactly what is required, and in the impact that it could have. Again the Introduction chapter is a great place to grab the reader, and show how important the work is. If possible try and find something that has just happened in the news in the introduction that shows how important your work is. The Introduction chapter, at least, should be readable by all, and where, at the end of it, most readers would want to read on, as it sounds so interesting.
  20. Make your thesis a sandwich. With a good thesis, we open with the Introduction and close on the Conclusions. The bit in-between justifies what you have opened with and the conclusions should show what you have uncovered to justify your argument. The same goes for each chapter, where the introduction (half a page, typically) shows what you're going to tell them, and the conclusion confirms it. Do not make conclusion into a summary, as they reader has no time to read summaries, and just wants you to conclude the most important things that go forward (and so they can dump all the other things that you covered). If possible say why you are not taking some things forward in the conclusions (and justify using the work of others, if possible). 
  21. Don't just pick without reviewing and justifying. There is no justification in a thesis for picking something just because it is easy to get. If possible all the things that are selected have at least been reviewed, and a sensible solution is selected (and justified). Try always to select a few competing methods and tools and put them against each others.
  22. Validate before Evaluate. You won't believe the number of Vivas that I've done where I've asked if they validated their system or software before they went onto evaluating it. So "How do you it takes 5 milliseconds to get from here to there?, the wrong answer is "... because the package said it was 5milliseconds". Good experimenters will do "fag-packet" calculations, to estimate things and know the limits of what they expect. I always like to see validation tests within the test data, so that the researcher knows that their system is working correctly. There's nothing work in finding there is a bug in your results, after you have published them ... so always have a sanity check.
  23. Get that scientific method. There are so many occasions in a thesis where you have no idea what a graph is telling you, as the axis' are not numbered properly, or where they are  poorly scaled. If the variation is between 990 and 1000, don't draw a graph which goes from 0 to 1000. Work out what the graph is trying to say, and pick the graph type (eg pie chart to show significant of one method against another) to show this.
  24. Must be based on a method and be repeatable. There must be a method in the processes used, and designed in a scientific way. Along with this the thesis should outline the procedure in a repeatable way, so that someone else can perform the same evaluation and get the same results. So candidates should always say to themselves... "Is there enough information for someone to build the artifact?", "Is there enough information to repeat the experiment?", and "Do I have the data that the examiner can look at, in order to verify the evaluation?"
  25. Evaluate your method against others. The standard method to show a contribution is to take your method, and evaluate it against other competing methods. The best approach is to use the best competing method and show an improvement. This can sometimes be difficult, so, at least, there is an evaluation against other methods. Showing an improvement is obviously a good thing, but there is often nothing wrong with an evaluation which shows a negative impact, especially if it is backed-up with a strong critical appraisal.
Oh, I stuck to 25, but there's a few more:
  1. Be fair and honest with your experiments. Often an experimental procedure is selected to benefit your own method. If possible be fair on all the methods and do not bias your approach to your one. It does no harm to show weaknesses and downsides to your own contribution, as it gives you a chance to critically appraise and show how future work could improve things. Your experimental procedure and the associated data collection should be repeatable and verifiable, so don't delete that data you have gathered.
  2. If possible, know your examination team. While the thesis should stand-alone you should also know your examination team before the Viva, so avoid patronising them with background theory which they know inside-out, or provide some background which might help the examiners to understand the area. Often an examiner, as part of the Viva, will give advice on moving things between the core material and appendices, in order to address the target audience for the thesis.
  3. Show that you are now an expert in your area. People expect those with a PhD to be an expert in the area of study, so make sure you know your core principles in the subject area. If you are doing a cryptography PhD read around the subject, and know the core principles of the most important methods. For me, anyone doing a PhD in electrical engineering, for example, should know Ohms Law, and the same should go for other subjects.
  4. Use appendices. Many PhD thesis' are full of material that is irrelevant to the many arguments, and writers are often too sensitive about removing material, but irrelevant material can often be placed better in an appendix.
  5. Quality is better than quantity. Some of the best thesis I read have been relatively short and sharp, but where the quality is high. A good eye for moving material in appendices is important, and helps the examiner. For some reason, candidates like to produce thick thesis, and they think that the more pages there, the better the material. This is often the opposite, and a thesis written with self-contained papers are often the best in their presentation.
  6. Define published work. A key part of PhD study is the dissemination of the work, especially with peer reviewed. The examiner often needs to know what has been published.
  7. Watch those unreliable references. In a PhD thesis, the references should be credible and verifiable references, and references to industry-focused white papers or general Web pages cannot be trusted providing credible viewpoints.
  8. Look for small-scale to large-scale experiments. A good researcher will often start small scale and prove principle, and then look for a large-scale experiment. The sign of small experiments, along with a large-scale experiment which properly evaluates the methods presented, is a good sign of a strong research ethos.
  9. Leave the Introduction and Conclusions to last, and then do the Abstract, and finally the title. You will know the full scope of you work once you have done the main chapters, so leave the Introduction and Conclusions to the end, and writing them together, with an opening statement and a concluding answer. The abstract then distils the whole of the thesis and pick a title that then reflect this (and that you are happy with).
  10. Few abstracts are actually any good in first draft. For some reason most PhD students struggle to write an abstract, and often it is written more as an introduction rather than a distilled version of the thesis. Remember that the abstract is the first thing that the reader reads, so if it is not focused on presenting the whole of the thesis, you have missed an opportunity to get the reader on your side. If possible an abstract should be a page in length, and outline the problem, the contribution, the most significant methods, the thing that has been designed/modeled, what has been evaluated, and what the most significant result is.
  11. Conclusions should conclude the whole thesis. Often the thesis just verifies aims and shows the significance of the results, but it should also recap the key parts of the literature and the other chapters.
  12. Mind those commas. Commas seem to be a dying bread, but are there to help speak directly to the reader. Try and read out loud, and if there's a slight pause, add a comma.
  13. End on a high! Don't spoil your thesis, by adding another chapter after the main contribution. Leave the reader on a high, and get them into the Conclusions, and leave the stage. I've read a few thesis' where the last chapter is a real let-down, and contibutes very little to the overall focus of the work. If you want, put your lovely new models in an appendix, and refer them in the main chapters, but try and finish the main chapters with the answer to the question posed at the start. The last dot of the last main chapter cements the argument, so don't run on into something else that you just happen to be looking at, and just feel your thesis isn't thick enough yet.
In the Viva:
  1. Be ready to defend, up to a point. You are unlikely to ever win with a debate with the External Examiner, as they typically have the experience to know when they are right. The Examiner does want to see you putting up arguments against theirs, and not bend. A strategy is often to debate the case, and try different routes of explanation, but then to take on their advice for any changes that would be required.
  2. Draw it out and keep it simple. Drawing diagrams and abstracting is a great way to explain your ideas, so wherever possible try to draw an abstraction to show a key point. Try not to overcomplex things, as they examiner is often looking for you to article complex ideas in a simple and understandable way.
  3. The simplest things are often the most difficult to explain. Many candidates go into a Viva thinking they will get probed on the complex areas of their work, but end up having to justify an extremely simple concept, that they have taken for granted. An examiner can often spot a weakness in some fundamental areas and probe around that, in order to see how the candidate thinks through a problem. So candidates should also try and be well versed on the fundamentals areas, especially when it involves maths.
  4. Know your examiners. Every examiner is different, and they have their own style. Some go from page to page, others read generally around significant parts of the work. They will generally have expertise in certain areas, so try and understand their motivations in their research, and some of their specialities, as they are likely to draw on these for questions.
  5. Don't leave it too long for the Viva. The best time for a Viva is straight after you've written your thesis, so try and don't leave it too long for the Viva, as you will forget a few things.
  6. Stay calm and enjoy. It is your opportunity to lock horns with an expert in their field, so enjoy it, as you'll probably never have the chance to do something like this in your career.

Conclusions

A PhD is a long road, and you learn along that road. The end result should setup you up for the even longer road ahead, but you now have all the tools to be ready for a career in research. None of us truly knows the formula for a successful PhD, but the methods applied by examiners and supervisors have stood the test of time, and do actually result in something that can contribute to the body of science. Remember that you are:
Standing on the shoulders of giants
a key thing is knowing who's shoulders you are standing on, and help the others who could stand on your shoulders. Enjoy your time!

And finally...

For a bit of advice, have a look at Ralph Merkle's time. He invented key exchange while an undergraduate, but his professor rejected his ideas because he didn't articulate them properly, and Ralph then tried to publish a paper on it, but it was rejected because he had no literature in the paper 
So, try and write well ... and perfect the art of speaking directly to the reader, and also follow the rules of research that have been laid down over the centuries, and you are half way there.

No comments:

Post a Comment