Step 1 of 5
Visual Mapping Principles
Choosing a chart is driven directly by your research goal and variable types.
Research 101 Modules
Presenting Your Data
A picture is worth a thousand words. Learn how to map data features to appropriate charts (bar, scatter, line, box plot) and avoid visual distortion.
Module complete. It is checked off in your course progress.
Chart Selector
- Goal
- Data
- Chart
Goal
What is the primary goal of your visualisation?
The goal of your visualisation is the most important first decision. Different goals require fundamentally different chart types, and choosing the wrong type can misrepresent your findings or make them harder to understand (Wilke, 2019).
Groups
How many categorical grouping variables do you have?
The number of grouping dimensions determines whether a standard bar chart (one grouping variable) or a grouped/clustered bar chart (two grouping variables) is most appropriate. More than two grouping dimensions typically require small multiples or a heatmap (Wilke, 2019).
Variables
Are you showing the distribution of one variable, or comparing distributions across groups?
A histogram shows the full distribution of a single continuous variable. A box plot compares summary distributions across two or more groups. For small samples, overlaying individual data points is strongly recommended (Wilke, 2019).
Categories
How many categories are you showing as parts of the whole?
Pie charts work only with a very small number of clearly distinct categories (2-4). With more categories, or when you also need to show totals, a stacked bar chart is more appropriate (Few, 2012; Wilke, 2019).
Comparison
Bar Chart
Also known as: column chart (vertical orientation)
Displays the magnitude of a continuous variable (e.g. mean, count, frequency) for each category using proportional bars. Bars can be oriented vertically (column chart) or horizontally. Each bar represents one discrete category.
Data type
Categorical (nominal or ordinal)
Variables
1 categorical + 1 continuous
Time axis required
No
Do
- Start the y-axis at zero — the eye judges bar length, not position, so a truncated axis distorts comparisons (Tufte, 1983)
- Order bars by value (descending) unless a natural category order exists
- Use horizontal bars when category labels are long or there are many categories
- Label axes clearly with units
Don't
- Do not use 3D bars — they distort proportions and add no information (Wilke, 2019)
- Do not use too many colors — one color (or one per group) is sufficient
- Do not use a bar chart for continuous data — use a histogram instead
- Avoid excessive gridlines or decorative elements (Tufte, 1983)
Example in practice
A researcher compares mean research anxiety scores across four lab types (wet lab, dry lab, field, computational). A bar chart with error bars (showing SD or 95% CI) communicates group differences clearly.
Sources: Wilke, C.O. (2019). Fundamentals of Data Visualization. O'Reilly Media. | Tufte, E.R. (1983). The Visual Display of Quantitative Information. Graphics Press. | Few, S. (2012). Show Me the Numbers (2nd ed.). Analytics Press.
Comparison
Grouped Bar Chart
Also known as: clustered bar chart
Extends the bar chart to compare a continuous variable across combinations of two categorical variables. Bars are grouped (clustered) side-by-side, with one grouping variable defining the clusters and another defining the bar color/fill.
Data type
Two categorical variables + 1 continuous
Variables
2 categorical + 1 continuous
Time axis required
No
Do
- Limit to two or three groups per cluster — more becomes unreadable
- Use a clear, accessible color palette to distinguish groups
- Include a legend explaining the grouping variable
- Keep bar widths consistent within and across clusters
Don't
- Do not use more than 3-4 clusters — consider a faceted (small multiples) approach instead (Wilke, 2019)
- Do not use patterns or textures as the only differentiator (accessibility issue)
- Do not omit a legend
- Avoid 3D effects
Example in practice
A researcher compares mean publication rates across three mentorship types (none, near-peer, faculty), separately for STEM and non-STEM students. A grouped bar chart with two bar colors shows both dimensions simultaneously.
Sources: Wilke, C.O. (2019). Fundamentals of Data Visualization. O'Reilly Media. | Few, S. (2012). Show Me the Numbers (2nd ed.). Analytics Press.
Trend over time
Line Chart
Also known as: line graph
Displays how a continuous variable changes over a continuous dimension, typically time. Points are connected by lines, emphasising continuity and trajectory. Particularly effective for showing trends, rates of change, and multiple time series simultaneously.
Data type
Continuous (time on x-axis)
Variables
Time + 1 or more continuous outcomes
Time axis required
Yes
Do
- Connect data points only when the x-axis is continuous (e.g. time) — never connect unordered categories
- Start the y-axis at a value that shows meaningful change without distortion
- Label multiple lines directly on the chart rather than using a legend where possible (Wilke, 2019)
- Use at least 5 data points to justify a line — fewer points may be better shown as a bar chart
Don't
- Do not use a line chart for categorical x-axis data — use a bar chart instead
- Do not include more than 5-6 lines without strong visual differentiation
- Avoid dual y-axes where possible; they can mislead readers about relationships between series
- Do not omit uncertainty (error bands or confidence intervals) for key findings
Example in practice
A researcher tracks student research self-efficacy scores monthly over a 6-month mentorship programme. A line chart clearly shows the trajectory of change, including any plateaus or dips.
Sources: Wilke, C.O. (2019). Fundamentals of Data Visualization. O'Reilly Media. | Tufte, E.R. (1983). The Visual Display of Quantitative Information. Graphics Press.
Relationship
Scatter Plot
Also known as: scatter graph, XY plot
Plots each observation as a point in a two-dimensional space defined by two continuous variables. Used to visualise the direction, strength, and form of a relationship (correlation) between variables, and to reveal clusters, outliers, and non-linear patterns.
Data type
Two continuous variables
Variables
2 continuous (optional 3rd as color/size)
Time axis required
No
Do
- Add a regression/trend line when you want to show the direction and fit of the relationship
- Use transparency (alpha) when many points overlap to reveal density
- Label notable outliers directly on the plot
- Keep axis scales that reflect the actual data range — do not force axes to zero for continuous ratio data if doing so wastes space (Wilke, 2019)
Don't
- Do not connect scatter plot points with lines unless observations are ordered in a meaningful sequence
- Do not overplot — use jittering, transparency, or binned hexplots for large datasets
- Do not imply causation from a correlation shown in a scatter plot
- Avoid bubble charts unless the third variable (bubble size) is essential and clearly labelled
Example in practice
A researcher plots hours of mentorship received (x-axis) against research output score (y-axis) for 90 students, adding a fitted regression line and colouring points by discipline to examine whether the relationship holds across fields.
Sources: Wilke, C.O. (2019). Fundamentals of Data Visualization. O'Reilly Media. | UC Berkeley Library Data Visualization Guide (2024). guides.lib.berkeley.edu/data-visualization.
Distribution
Histogram
Not the same as a bar chart — bars represent continuous bins, not categories
Displays the distribution of a single continuous variable by grouping values into contiguous bins and showing the frequency or density of observations in each bin. Reveals the shape (normal, skewed, bimodal), spread, and presence of outliers in the data.
Data type
One continuous variable
Variables
1 continuous
Time axis required
No
Do
- Choose bin width thoughtfully — too few bins hides patterns; too many creates noise (Wilke, 2019)
- Label the x-axis with the variable name and units
- Indicate whether the y-axis shows frequency (count) or density (proportion)
- Superimpose a normal curve if testing the normality assumption visually
Don't
- Do not use a histogram for categorical variables — use a bar chart
- Do not leave bars with gaps between them — gaps imply discrete categories (unlike a bar chart)
- Avoid using histograms for very small samples (n < 20) — a dot plot or box plot may be clearer
- Do not choose bin width arbitrarily without trying multiple options
Example in practice
Before running a t-test, a researcher plots a histogram of reaction time data from 60 participants to check visually whether the data appear approximately normally distributed and whether outliers are present.
Sources: Wilke, C.O. (2019). Fundamentals of Data Visualization. O'Reilly Media. | UNT Health Science Center Library. Types of Charts. libguides.unthsc.edu/data-visualization.
Distribution
Box Plot
Also known as: box-and-whisker plot
Summarises the distribution of a continuous variable using five statistics: minimum, first quartile (Q1), median, third quartile (Q3), and maximum, with outliers plotted individually. Particularly useful for comparing distributions across multiple groups simultaneously.
Data type
Continuous variable across groups
Variables
1 continuous + 1 categorical (groups)
Time axis required
No
Do
- Plot individual data points overlaid on the box when n is small (n < 50), so readers can see the raw data (Wilke, 2019)
- Label each box with the group name on the x-axis
- Report sample size per group either on the plot or in the caption
- Include a caption explaining how outliers are defined (typically >1.5 IQR from Q1 or Q3)
Don't
- Do not use a box plot as the only visualisation for small samples — the box implies more data than may exist
- Do not omit outliers — they carry important information
- Avoid violin plots for small samples (n < ~50) as the smoothing becomes unreliable
- Do not connect medians across groups with a line — this implies a continuous trend
Example in practice
A researcher compares the distribution of exam scores across three teaching methods. Box plots for each group simultaneously display the median, spread, and outliers, making group differences and variability visible.
Sources: Wilke, C.O. (2019). Fundamentals of Data Visualization. O'Reilly Media. | UNT Health Science Center Library. Types of Charts. libguides.unthsc.edu.
Part-to-whole
Pie Chart
Use only for part-to-whole relationships with very few categories
Displays the proportional contribution of each category to a whole by dividing a circle into wedges. The angle of each wedge is proportional to the category's share. Appropriate only when the focus is on part-to-whole relationships and there are very few (2-4) categories.
Data type
Categorical (nominal), few categories
Variables
1 categorical + proportions summing to 100%
Time axis required
No
Do
- Use only when the parts genuinely add up to a meaningful whole (e.g. percentages of a total)
- Limit to 2-4 categories — more wedges become impossible to compare accurately
- Label each wedge directly with its percentage and category name
- Sort wedges from largest to smallest, starting at 12 o'clock
Don't
- Do not use a pie chart when precise comparison between categories is needed — use a bar chart instead (Few, 2012)
- Never use 3D pie charts — the perspective distorts relative areas substantially (Tufte, 1983)
- Do not use a pie chart when any category is very small — thin wedges are unreadable
- Avoid donut charts when the central hole provides no additional information
Example in practice
A researcher reports that 45% of study participants were from STEM, 30% from social sciences, and 25% from humanities. A pie chart (3 wedges, directly labelled) shows the overall composition of the sample.
Sources: Few, S. (2012). Show Me the Numbers (2nd ed.). Analytics Press. | Wilke, C.O. (2019). Fundamentals of Data Visualization. O'Reilly Media. | Tufte, E.R. (1983). The Visual Display of Quantitative Information.
Part-to-whole
Stacked Bar Chart
Consider grouped bar or 100% stacked bar as alternatives
Extends the bar chart by dividing each bar into stacked segments, each representing a sub-category's contribution to the total. Shows both the total magnitude and the part-to-whole composition simultaneously. The 100% stacked variant shows proportional composition only.
Data type
Categorical variable with sub-categories + continuous
Variables
2 categorical + 1 continuous
Time axis required
Optional
Do
- Place the most important sub-category at the bottom, anchored to the baseline, so it can be compared accurately across bars
- Use a 100% stacked bar when the focus is on proportions rather than absolute values
- Use clearly distinguishable, accessible colors for each segment
- Include a legend and direct labels where possible
Don't
- Do not use stacked bars when precise comparison of middle or top segments is needed — only the bottom segment is easy to compare (Wilke, 2019)
- Do not stack more than 3-4 sub-categories — beyond this, patterns become impossible to read
- Avoid using both stacked totals and proportions in the same chart without clarity
- Do not use patterns or textures as segment fills
Example in practice
A researcher shows the composition of research output types (first-author papers, co-authored papers, conference abstracts) across three cohort years. A stacked bar anchors the first-author category at the base for easy comparison.
Sources: Wilke, C.O. (2019). Fundamentals of Data Visualization. O'Reilly Media. | Few, S. (2012). Show Me the Numbers (2nd ed.). Analytics Press.