The Methods section explains exactly what you did and why you did it that way. I think of it like a recipe. Someone with the necessary data and resources should be able to read the section and understand how to reproduce the analysis. That goal is called reproducibility. It does not mean another researcher will always get an identical result in every situation. It means you gave enough detail for them to repeat the steps instead of guessing what you did.
The Basic Parts of a Methods Section
Before writing full paragraphs, I would make a checklist of the information the reader needs:
- Where did the data come from?
- Who or what was included?
- Who or what was excluded?
- What was the outcome?
- What variables were examined?
- How were those variables categorized?
- What statistical tests or models were used?
- How was missing information handled?
The exact order may change depending on the study, but none of these details should be hidden. If a choice affected who entered the analysis or how a result was calculated, explain it.
Define the Population Precisely
Saying that a study included “people with cancer” may not be specific enough. Did it include children and adults? Did everyone have a current diagnosis? Were they receiving treatment, waiting for treatment, or finished with treatment? Using my cancer-treatment paper as an example, the target population ultimately had to be described as adults with cancer who were receiving or scheduled to receive treatment during the pandemic. That wording tells the reader who the results actually apply to.
This is also where inclusion and exclusion rules belong. If an age limit, missing answer, treatment status, or another requirement removed someone from the analysis, say so. The final sample should not appear without an explanation of how it was formed.
Define the Outcome
The Methods section is where the reader gets the technical definition of the main outcome. A vague outcome can make the entire paper hard to interpret.
For example, I originally used the word change without specifying what it meant. That was a problem because a treatment change could mean several different things. It could mean switching treatments, delaying one, canceling one, or something else. I changed the wording to disruption, but that still had to be defined. In the paper, a change, delay, or cancellation was combined into one binary outcome. Binary simply means the final outcome has two categories: a person either reported a treatment disruption or did not. A reader should know that definition before reaching the Results section, because otherwise the numbers there are hard to interpret correctly.
Explain How Variables Were Coded
Data do not always enter an analysis in the same form in which they were collected. Age may begin as a number and then be separated into age groups. Smoking responses may become never, former, or current. A yes-or-no outcome may become 1 or 0 for analysis. Explain each of those decisions. The appendix of my paper mapped the original survey questions to the final analytical categories. That kind of mapping is useful because the reader can see exactly how an answer in the survey became a variable in the model.
If there are too many definitions for the main Methods section, an appendix can hold the full coding table. The main section should still summarize the important choices and point the reader to the details.
Justify the Covariates
If you are doing data analysis, make sure to justify how you select confounders.
The scientific terms here are confounders or covariates. These are other characteristics that may be related to the group being studied, the outcome, or both. They matter because an adjusted model is trying to estimate the association between one variable and the outcome while accounting for those other characteristics. Do not put every available variable into a model just because it is in the dataset. Explain why each one is relevant. That reason might come from prior research, knowledge of the subject, or the design of the study. The cancer-treatment paper states that its sociodemographic and health variables were selected because prior literature suggested they might be associated with delayed cancer treatment. That explanation tells the reader why those variables were included instead of making the model look like an unexplained list.
Explain the Analysis in Order
If you are planning to run data analysis in software like RStudio, it helps to describe the analysis in the same order you actually carried it out. A beginner-friendly analysis plan can follow three steps:
- Describe the sample.
- Estimate how common the outcome is in each group.
- Run a regression model to examine adjusted associations.
This is the progression I used. The order lets the reader first understand who was studied, then see the descriptive results, and finally understand what the adjusted model tested. Name the software or programming language used for the analysis. Also name important packages and main functions when they matter. However, a list of function names is not an explanation. The reader should still be able to tell what each part of the code accomplished, such as applying survey weights, calculating prevalence, or fitting a regression model.
If you are unsure about the code, there are plenty of YouTube tutorials that teach the type of coding I did (which was very basic). Claude and ChatGPT are also very good at writing simple code like this. They are also great at debugging your code if you are having trouble.
Explain Missing Data
Missing answers can change an analysis, so report how they were handled. Were observations with missing values excluded? How many observations were lost? Did different analyses use different sample sizes? My paper included a missingness table showing how much data were absent for each covariate. That lets the reader see whether one variable had many more missing responses than another. Without that information, a smaller analysis sample can be easy to miss.
Be Specific and Consistent
Some key takeaways I had were to be very specific, make sure you’re using the same terminology throughout, and make sure you’re very precise in your wording. The Methods section is probably where this matters most. If the outcome has one name in the abstract and a different name in the Methods, the reader may wonder whether the paper is discussing two different measures. Pick the exact term, define it, and keep using it.
The Recipe Test
Give the Methods section to someone who understands the field but did not work on the project. Can they identify the exact dataset and sample? Can they explain what every important variable meant? Can they tell which analyses were performed and in what order?
