How does one balance between working on a research problem and writing up the results?

Question

I currently have two months before I have to submit my master's thesis (its a one-year program in the UK) and just finished data collection. Analysing the data is a time consuming process and I'm starting to worry that I don't have too much time to write my dissertation. Will it be better to prioritize the analysis (and actually get some reportable results) or to get cracking on the writing once and for all (since our final grade would ultimately be based on that)? I feel like an even split between the two would be best but its hard trying to find the right balance. Both require very different mindsets. When writing I need to be in a quiet place where I can slowly churn out words and read (or skim) papers I come across. Analysing data happens in a noisy lab where distractions are everywhere but ultimately work gets done at the end of the day.

At this stage, should I just focus on getting the analysis done with (which could take weeks since its highly exploratory in nature) or should I try to somehow come up with a strategy to accomplish both?

Answer 1

Do everything you can to reduce the exploratory nature of the data analysis.

Plan exactly what results you are going to need for your manuscript, what the tables will be, what the figures will be, how they will be formatted. Then when you get to the data analysis, you can be extremely focused, concentrating only on what's essential for your work.

Answer 2

If you are doing your analysis with R, you can kind of do both at the same time using knitr. This package allows to mix code, the code output, and normal text in the same file. There is also great control in what is show. For instance, in the final thesis all the code could be hidden. In effect, you would be doing the exploratory analysis whilst writing down how it was done and the description of the results. Since it is plain text documents, you can complement it with version control and a free online hosting (github or bitbucket), and have backups of the analysis. Even though I don't use it on a regular basis, RStudio would also be of great help here.

Knitr offers markdown and LaTeX. The former is very easy to use and it will take you a about a hour to learn. LaTeX which is more rich in functionality, but takes way longer to learn. Using pandoc, the final document could also be converted to .doc.

But even if you are not using R, write down what (how) you are doing, and summaries of results. This will save time when doing the write-up.

Now to answer your original question:

At this stage, should I just focus on getting the analysis done with (which could take weeks since its highly exploratory in nature) or should I try to somehow come up with a strategy to accomplish both?

Write down what are questions you are trying to answer, and which steps are needed to answer them. Then access which would be the minimum required for a Thesis - your supervisor could also help you with this. This will allow you to focus on small chunks of your thesis (more manageable and preferably independent), and the end of each access how much time you have left for the writing of conclusions/introduction or whether you can or should tackle another one. It is also ok, to leave out results from the final thesis if there is not time to explain those in-depth.

Personally, I would err on the side of less analysis, but well done rather than a hodgepodge of half-baked attempts. Keep in mind that the goal here is to show that you have done research and acquired new knowledge.

Answer 3

Yes, you have to do both. You might be able to write some parts of the thesis in advance, but the analysis must precede writing about your results.

Answer 4

I write as I go. I begin the introduction and methods as I begin working on a new project, I write results as (or even before) I make plots. Through the process of collecting data, I get new ideas that I want to capture in a rough draft of the discussion. Writing up even rough results makes me think more clearly and helps me determine what data I need to gather and plot next. Through this process, I throw away about as much text as I end up with. Writing a good draft requires writing a bad draft.

If I did all the analysis, then wrote the bad draft, then wrote the good draft, the bad draft would only help with the writing. By writing every bit of the bad draft as soon as possible, the bad draft can help myself and my collaborators think more clearly about how to improve the analysis. Writing as I go lets me use the bad draft not just to help with the writing, but with the thinking.

Answer 5

The ideal way to perform research is to begin with a set of questions. If new questions are generated based on the data, that's all well and good, but unless the findings actually invalidate the questions themselves—which occasionally happens—you'll be helping yourself by solidifying your questions before you start.

To try to bring that to your specific question, given that you have such a limited timeframe, focus all your analysis time on answering the question you were posing at the outset. Once that's done, write until you've completely written up that analysis. If, at that point, you have other interesting analyses you want to perform, you can do the analysis and write it up.

The only exception I would list here is the case where you entire analysis failed, but something else very interesting (i.e., publishable) came up that is more exciting than your first question. I only mention this because this happened to a colleague of mine; his project failed, but he found something else fascinating, and the main focus of his thesis was the second thing. That case aside, though, definitely focus on your main question at hand.