What I Learnt This Week (#41 – 2016, Paper Writing Edition)

This week has been crazy with two paper submissions. I will tell more about them later, maybe. During this week, I have made a couple of realizations regarding paper writing. I thought it was cool to put up this small list of must-knows and tips, from my short experience.

Data Gathering Tips

Write a script that relates the main interactions between the subject and the researcher. Be sure to comply with the ethical recommendations of your institution. Usually they refer to ask for permission and anonymization of the data before publication.

Be extremely cautious with possible data leaks. Handle the documents with care and think twice before making copies of the originals. Keep them as long as needed to comply with the data protection schemes you have to abide by.

We all know that our experiments take time to prepare and that we all want to get the results we predict, or at least them to be usable, but cherrypicking the data to show only the results you want is not an option. If the results don’t seem to be what was expected, just don’t despair. You are not the first neither the last to whom that has happened. That is inherent to science and discovery.

The statistics pack

I realized that there are many things that you will most likely need to do all the time when writing a paper, so maybe it is good that you refresh your statistics and make sure you know how to do

  • Mean
  • Median
  • Mode
  • Variance
  • Standard Deviation
  • Standard Error
  • T-Test

And most likely many more that are far less usual.

Graphing tools

Whatever tool you use (Excel, R…) just know the basic graphs that you will need to produce. This means, usually:

  • Line graphs
  • Bar graphs

Less, usually, but still appear sometimes:

  • Polar graphs
  • Pie graphs

And furthermore, learn how to add the error bars or envelopes. They usually help a lot to understand the relative changes of the data samples.

Drawing and Scheming tools

It is always cool to include graphics to explain the basic concepts of the paper, so it is cool to have a diagramming tool like Dia in the toolbox. I tend to avoid using PowerPoint and the like to do diagrams.


Latex is a hard nut to crack, but it is worth to get to it as soon as possible. This is a good place to start, even though it looks a little bit dry. Problems I usually come across:

  • Tables: inner lines, titles, formatting of individual cells, full width of the table or of each of the individual cells.
  • Images: when the format is column-based, they always bring problems if you want the image to use the width of more than one column or the full width.
  • References: always some rogue reference appears as a couple of question marks, check them.

As a general rule, “study” the template if it is given to you before start modifying it. Many venues require to follow certain rules. Use some writing environment such as TeXnicCenter. It really eases work.


Usually papers are teamwork. As a recommendation, learnt through mistakes, it is good that is clean:

  • Which person is responsible of each aspect of the paper (data gathering, data analysis, writing, referencing, proofreading…)
  • In which order the authors are going to be written in the final edition and submission.

Data Analysis Tips

I always try to process the data through the means of software, as it lowers the chances of a mistake. When I have multidimensional data, I tend to craft an application (Python, Java or C#, usually) that transforms the raw data from the logs into a database to which I can make queries.

In the long term, this has given me very good results, as you can extract with ease much more meaningful data. The data takes longer to model and the software to program, but at the end this is much compensated by the ease to make queries to fit the needs and story line of the paper.

Writing a paper is not an easy task, and requires tools and skills. This is only a very small set of notes made out of a very short experience.

Leave a Reply

This site uses Akismet to reduce spam. Learn how your comment data is processed.