Data Analysis in the Research Cycle

• Data is often the first step in the research cycle, as it is required to form a hypothesis.
• The process you use to manage and analyse data is a workflow.
• You should always record your workflow so it is reproducible and understandable by yourself and others.

1 Data in the Cycle of Science

The cycle of science is often presented as:

1. formulate a hypothesis
2. test the hypothesis
   1. gather data
   2. analyse data
3. update the hypothesis
4. repeat until done

Figure 1.1: The cycle of science

However, it may be difficult to form a hypothesis without having first made some observations of the system under study, or a similar system. These observations are data. The scientific cycle has always started - even if unacknowledged - with data collection and analysis. For instance, Darwin and Wallace both made long journeys with many observations before they formulated their hypotheses of descent with selection.

In the curent age of data-intensive research, researchers who do not identify as data scientists are having to work with large, complex datasets. The focus of modern research can be very much on open-minded data analysis to formulate hypotheses as a first step in the cycle. With the ever-increasing amount of freely-available public research data, we can get started on formulating hypotheses more easily than ever before.

It is therefore especially important that your data exploration is recorded, replicable, and systematic.

2 Pipelines and Workflows, Oh My!

We can distinguish between two terms that are often used interchangeably: pipelines and workflows.

• A Pipeline is a set of instructions, often run by a computer, that takes a dataset and pipes it through a process or series of processes, until a result emerges from the end.

• A Workflow is a process that researchers use to investigate a problem.

A workflow may involve using one or more pipelines, but includes things like exploring your data, forming hypotheses, writing code, and interpreting your results.

In these workshops, we are concerned with the complete research workflow.

3 Explore, Refine, Produce (ERP)

One way to think about your workflow of data analysis is as an “Explore, Refine, Produce” cycle¹.

Figure 3.1: The Explore, Refine, Produce cycle. Reproduced from Stoudt et al. (2021), doi:10.1371/journal.pcbi.1008770.g001

In the Explore phase you process and interrogate your data to identify potential solutions. This phase is centred on you as the researcher, as you dig deep into your data.

In the Refine phase, you narrow down your focus to the most promising approaches. This phase is about how you, and your team, work towards a satisfying solution.

The Produce phase is really a companion to the Explore and Refine phases, and is about how you communicate your results to the wider community.

You should always be recording your work and preparing it for future consumption by yourself, or by others.

An essential skill for this is data management

---

1. Stoudt et al. (2021) “Principles for data analysis workflows” PLoS Comp. Biol. doi:10.1371/journal.pcbi.1008770.g001↩︎