Data Visualization Program

As a cross-cutting program, we work across the many disease areas, study types, and data types that the QSU supports. We are developing an internal team with particular expertise in visual thinking, drawing on staff with various biostatistical specialties and experience with various clinical settings. 

Visual thinking and data visualization are applicable to all study types, although we have found it particularly useful for:

·       observational studies where the causal chain may be hard to establish

·       studies using pragmatic or real-world datasets such as EHRs, market data, or registries

·       studies with complex or novel designs 

·       studies grappling with various forms of bias (selection, informed censoring, collider, etc.) 

·       studies that intend to communicate results not only to other researchers but also to other audiences such as patients, administrators, or the public.

If you are interested in working with the Data Visualization Program in your research, please complete a QSU project initiation form (QSU-partnered entities only.)

Our expertise in visual thinking supports research quality across all phases of a study:

Setting the research question and study aims

This pivotal stage of research requires attention, specificity, and consensus. For some funded studies the aims may have already been established, but often they are not fully written or require revision. At this stage, we lead a collaborative visual thinking process to establish specific aims, based on a clear DAG, logic model, or other conceptual study diagram. Our graphical skills help these to be clear, concise, and easily interpretable, and we find that these foundational diagrams can both support substantive discussions early on and speed various processes later.

Study design

Through a visual design process, we arrive at a study design yielding the highest possible level of evidence given the aims, data, and any other constraints. One deliverable from this process is a strong study design diagram; while study diagrams are common, our design process can make them clearer, more informative, and easier to process visually.

In the process of study design, we also generate mock figures of results and lead a process of interpretation. We consider mock figures with various study outcomes (including a null result) and discuss and record how we would interpret them. These are invaluable once true results are available.

Data management

While data is being collected, transferred, or accessed for a study, many unexpected data quality issues can arise. We use visual means such as well-designed graphs and diagrams to efficiently describe these issues and propose solutions. 

Statistical analysis

Often the first stage in statistical analysis is to assess the data. We engage in a visual process to diagnose data issues such as missingness, distributions of critical variables, collinearity, and so on. 

In the end our statistical analysis plans (SAPs) include all the visuals developed so far – aims, conceptual diagram, study design diagram, etc. -- to support statistical approaches and decisions.

Interpretation of results

At this point we complete the mock graphs from the study design process and use them to draw appropriate conclusions, including some draft language for results sections. 

Publications

We work with researchers on figures for publication, applying our data visualization expertise, statistical knowledge, and awareness of journal requirements. As appropriate, we use a variety of lesser-known graph types such as Marimekko charts, Sankey diagrams, raincloud plots, upset plots, and Hypothetical Outcome Plots (HOPs), each with their own strengths in certain contexts. We find that many journals are more accepting of innovative graph types than we might expect.

We also specialize in the representation of uncertainty in publication figures, based on recent research on how humans map uncertainty onto visual features of a graph.

Other visualizations

Designing a slide deck for conference presentation calls on different design principles than those for publication figures. We work with researchers to visually optimize their slides for presentation.

We also work with collaborators on other presentation modes, such as posters, visual abstracts, and key figures for social media. For each of these modes, design optimization is far more than cosmetic. Research shows that results presented visually using cognitive and graphical design principles have more reach and impact compared to others.

If you are interested in working with the Data Visualization Program in your research, please complete a QSU project initiation form (QSU-partnered entities only.

We offer advice on visual design for figures, graphs, posters, slides and the like. The sessions focus on:

·       how the content reflects the study aims and design, 

·       how the design can be more cognitively ergonomic, and 

·       how the design fits the intended venue, such as a slide presentation vs. a poster session.

We offer 1-2 office hours per week, by appointment only. If you are interested in an appointment, please complete a QSU project initiation form (QSU-partnered entities only.)

(Please note that we are software-agnostic and do not provide technical support or code for any program or graph type. We focus on concepts and design.)

Quarterly speaker series

We will have a range of speakers on data visualization, bringing perspectives such as:

·       Researchers developing new data visualization techniques relevant for clinical research

·       Researchers investigating how the visual system processes various graph types

·       Practitioners using visual thinking and visualization in their research

Our inaugural speaker was Dr. Will Stahl-Timmons (PhD MA AHEA), Information Designer for the British Medical Journal (BMJ), who discussed the BMJ’s process for creating and curating visual abstracts.