Some advice for your thesis

tips.md

Some advice for your thesis

Numbers, numbers, numbers

• Please, round all the numbers to a reasonable number of digits. In engineering, four significant digits are enough. And if you can compute 10 significant digits, you can be world-famous and earn $100.
• If you perform any experimental measurements, compute the confidence intervals too. alpha = 0.05 is a good default value for the significance level.
• If you have less than ca. 30 measurements, use only nonparametric statistics. If you have more than 30 samples, it is a good idea to check them for normality.
• Correlation does not imply causation
• Always specify units of measurements. siunitx is a great LaTeX package for this.

Do not do the same work twice

• When referencing a book or a paper, always specify the pages and the section/equation/figure/table/proposition/algorithm/... number. It will help you greatly, since all of us forget things that were a couple of months before---and you surely don't want to search through a 1000-page book only for this one proof. Examples:
  • \cite[(3.4), p.~412]{Bernstein1992} = Equation (3.4) on page 412
  • \cite[\S~3.1, pp.~81--84]{Sontag1998} = Section 3.1 on pages 81 to 84
  • \cite[Lemma~3.14, pp.~49f.]{Antoulas2005} = Lemma 3.14 on pages 49 and 50. 'pp. 49f.' means page 49 and the next page.
  • Quite often you will see something like 'pp. 20ff.', i.e. from the page 20 on. It's bad style, avoid this and always specify the exact range.
• A lot of people export a figure from e.g. MATLAB once and then include it into the document. Then, a week later, they discover some minor errors in the figure. And now they have to go to MATLAB, make this figure once again, configure the labels and axes, set the colours etc. Avoid this workflow! Instead, treat figures as secondary data and use primary data: Write a MATLAB script that does all these things for you and can create a figure from some data with the labels, axes and colours you want.
• The same goes for tables: Use a table generator that generates LaTeX code directly from numerical data. Not only will this make your job much easier, but also prevent typos and errors in the document.
• Should I tell you that you can also quickly write a TikZ code generator?
• TODO: Describe the workflow with aux-funs

Learn your tools of trade

• Git
• Python / MATLAB / C / C++ / Rust / ...
• LaTeX and BibTeX (DOI import)
• TikZ and TikZEdt
• Dash / Zeal

You and your advisor

• After some time into the project, you will often find yourself having a better knowledge of your thesis topic than your advisor. It is ok, since it is your thesis, not theirs. Just keep that in mind and be patient when explaining your recent findings.
• It is very helpful if you can chat with your advisor about some random things. Being focused on one thing can be harmful for your creativity.

Fighting laziness and procrastination

• Creativity is not steady. You will have great days on which you have a lot of new ideas and you will have totally boring days. So if you are having a meh day, do something different but still useful: Check spelling, polish document formatting, document your code, tidy up your desk, go for a walk etc.
• A lot of people I know (including myself) have the 'blank page' phobia, i.e. they look at their thesis, which consists only of placeholders and section titles, get depressed and go away. Editing an existing text, on the other hand, is much more productive. Hence, I recommend you to write down anything that comes to your mind. Just pretend you are William Faulkner and write, write, write, do not think. It is ok if it is totally informal -- the point is to have something you can edit.

Language

• Scientific English (or German or any other language) is not similar to the everyday language. You have to find out the commonly used collocations and memorise them. Therefore, read a lot of papers or textbooks and note not only the technical information, but also how the language is used.
• It does not matter if you write your thesis in American or British English; just stick to one flavour and use it consistently. Keep in mind, however, that a vast majority of the existing library functions are named using American English (color, optimize etc.).
• Be careful with the lingo, technical terms and definitions. For instance, covariance has different meanings in probability theory, computer science, vector calculus and physics.
• If your first language is German: Avoid ultra-long sentences like the plague! German allows you to be more specific in subordinate clauses, so a long German sentence will probably become convoluted and ambiguous if translated literally into English.
• Do not trust linguee.de. It often uses data from German web sites; and since the translators are mostly German, they can use an inappropriate Denglish term. Use Wikipedia instead: Just look for the article in German and jump to the same article in English. It gives you much more context!
• If you are not sure about how a word is pronounced, go to forvo.com.

Yep, good idea.