-
Notifications
You must be signed in to change notification settings - Fork 0
Writing your (first) paper
By Joseph Therriault
Here are some tips that I hope are helpful when it comes to writing your first paper. Maybe they can be applied to your next papers too. Starting to write is daunting but here I try to break it down step-by-step into something that is hopefully more manageable.
Usually, a paper will have 3-4 main text figures, but this can vary depending on the journal and how you want to show your results. This guide assumes you are done analyzing your data and have the “story” of your paper set (i.e. all your results, what order you will present them in, and how they tie together). The title for each figure should be how you explain the result in 1 sentence. Ideally, you want your figure to be understandable just by looking at it. Nonetheless, your figure legend will describe different elements of the figure and what they represent (i.e. “the dashed line indicates the threshold for tau positivity”; “Boxplots are presented for each biomarker, where lines indicate 95% confidence intervals and individual circles are data points that lie outside 95% confidence intervals”, etc). In the figure legend, if you used any kind of statistical model, write out all the covariates you used. If you are using acronyms, define all the acronyms in the figure legend.
Go back to previous papers from the lab and get the relevant study methods. For most studies, you will have a general description of the cohort, a brief description of how participants are diagnosed, a summary of our imaging methods, a summary of our fluid methods, and a statistical methods section. All are very important, but I think it’s very important that you understand your statistical methods section well, and that it is very clear.
Usually, the first paragraph of the results section will be about the demographics of your sample. Create a table summarizing all the relevant demographic, clinical and biomarker information for your sample. This will be Table 1. I like to use the Table1 package in R for this.
Then, each results paragraph is dedicated to one figure. If you don’t know where to start for the paragraph, something like “Figure 2 displays the relationship between X and Y” is good. The title of that figure will be something like “relationship between X and Y”, but you can refine it to better reflect the main messages of your paper too. Then describe the results in more detail, and report all relevant statistics in the same paragraph. Sometimes it can be boring and tedious to read a paragraph where every sentence is broken up by a lot of numbers and statistics. If you have a lot of statistics to report for a particular result, consider creating a Table to summarize them all together. Remember to include 95% CIs wherever appropriate, and avoid just reporting p values in isolation (i.e. also report effect size, beta estimate, or mean fold-change as appropriate).
This section is often intimidating, but if you follow a plan it isn’t so scary. The first paragraph should be a super brief summary of your paper. The first 1-2 sentences of your discussion should just summarize what you did. Then, summarize your keys findings in 2-3 sentences. Then one final sentence about the “big picture” implications of your work. This last sentence is often similar to the last sentence of your abstract. Reviewers and journal editors often jump to this paragraph of your discussion after looking through the abstract when evaluating your paper, so make sure it is very clear.
Second paragraph: summarize results of figure 1 (second paragraph of your results section) in 1-2 sentences. Then 3-4 sentences linking this finding to other studies in the field. Then, your final sentence can say something like “taken together, these results contribute to a framework in which XYZ”. The idea of the final sentence is that your paper, in combination with the other papers you cited, support a certain conclusion.
Repeat process above for all the main results in your study. If you have a hard time linking your study to other papers in the field, find a review paper written in the last 5 years on the topic (the more recent the better). Take the time to read the review carefully, taking notes on how your paper fits with the field, and keep track of all the references that will be relevant to your paper.
The last 2-3 paragraphs are the more general implications of your study. This is a good place to outline the high-level pertinence and impact of your paper. Follow the same format as the previous paragraphs: topic sentence, 4-5 sentences describing other studies / results (with references), then a summary sentence that links your paper and the other papers you mentioned.
Write your limitations paragraph. If you’re not sure, first look at other papers that did similar things and see if they apply. It’s usually good to mention limitations with respect to our study sample (i.e. people participating are individuals motivated to participate in a study of aging and AD and may not be demographically representative of the populations at risk for dementia). Some people like putting study strengths into this paragraph too.
Then write a final concluding sentence if you like (though some journals require it).
It is better to write the introduction towards the end of this process, because having written your discussion, you know your paper’s main message (i.e. what you will be introducing). Try to have 3-4 paragraphs at the most. A common mistake new students make is writing very long introductions where they try to include everything they know about the topic. But it’s important to remember you are writing for an expert audience, who already knows a lot about this field. Focus only on what is essential for your paper. What problem is being addressed? Why did you do this experiment? What is unknown and why is it important? To “build” the introduction, mention the studies that brought you to the question you are asking. If you like, in the last sentence of your last paragraph, you can include something like “here, we performed ABC in order to test XYZ”.
Depending on the journal, this can look very different, so at this point you need to have a good idea of where you want to submit. The abstract will need to outline and introduce the problem in 2-3 sentences, summarize the methods, summarize 1-2 key results, and have 1-2 sentences explaining what this shows for the field and why this paper is important. Writing an abstract is hard and you will probably go through many revisions, so don’t be too attached to it. But an abstract is also very important. Many researchers are very busy and may only read your abstract and skim through your figures. So make sure you are covering the most important elements of your paper.
This should fit onto one page (excluding references) and have 3 paragraphs. 1: Start with a sentence like: “We are pleased to submit our original article entitled Title (Maniscript ID: #) to the Journal of Great Results. Then provide a brief summary of the field, and identify the knowledge gap. 2: How your study fills the knowledge gap (2-3 sentences explaining key results). 3: Why your paper is important: something like “we believe our paper is suitable for publication in the Journal of Great Results because we identify XYZ as a key feature associated with disease progression”.
- Remember that editors and reviewers are busy people. Don’t make them have to scour through your paper to find something. Having things organized will make it easy to find the information they need. Sometimes, that also means writing the same thing multiple times in different sections so that people can’t miss it. It helps to include information in your figure legends like what covariates you had in the model so that it’s right in front of the reviewer as they are looking at the figure.
- Consider having a flowchart figure displaying how you arrived at your sample size (this can go to the supplementary materials). I find they help make your paper more clear, and when I review a paper I find them helpful.
- Try to avoid as many acronyms as possible. They make a paper really hard to understand for no reason - even for people in the same field but who specialize in something different. Try to reserve them for long clunky terms like MMSE, CDR-SoB, p-tau and so on. But no need to make “Grey matter” in GM and “hippocampus” into HC.
- Try to avoid a figure where you just show a map of the brain without having an accompanying scatterplot. Without the scatterplot, it’s hard to understand what the data actually look like. When I used to just show t-maps, my papers used to get rejected a lot more often.
- When you have a lot of results (i.e. reporting AUCs or group differences for multiple different biomarkers) it really helps to put all of that into a table that people can easily refer to. This can also make your results section easier to read.
- It’s ok to show the same thing multiple times in different ways. In fact, this is helpful: it shows your results are not driven by a specific analytical choice. Almost all journals allow supplementary materials. So when writing your results, it can often be good to have sentences like “we observed the same pattern of results when dichotomizing amyloid-PET status (supplementary Fig. 2).”.
- Avoid sentences like “our paper proves that XY” in favour of sentences like “our study provides evidence that XY”. Also, try to avoid causal language as much as possible, but if you really want to use it, be careful with your wording and use terms like “may cause”.
- If you can, try to cite many different research labs in your introduction and discussion. One of them is going to be your reviewer, and the reviewer is a human being. They will be a bit miffed that their hard work was ignored and another lab’s papers were cited over and over. A great example of this is idea is this paper’s first sentence (https://alz-journals.onlinelibrary.wiley.com/doi/full/10.1002/alz.13651) which cites 17 different publications that addressed a specific topic.
- After your first draft is done, sit down with someone senior from the lab (but less senior than Pedro) to go through your paper paragraph by paragraph. They will teach you little things about phrasing that will make your next papers easier to read (and to write).
If you have any questions, contact me (Joe).