Data Literacy and Visualisation
Domain 5 - Data Literacy and Visualisation. This domain tests whether you can read a chart critically, choose the right chart for a variable type, and compute and interpret basic summary statistics.
Core rules and concepts
- Variable types: nominal (unordered categories), ordinal (ordered categories), discrete numeric (counts), continuous numeric. Chart choice follows from the variable type, not from decoration.
- Chart matching: bar chart for comparing categories; histogram for the distribution of one continuous variable; line chart for time series and trends; scatter plot for the relationship between two numeric variables; box plot for spread and outliers; pie chart only for parts of a single whole with about five or fewer slices.
- Summary statistics: the mean is pulled by outliers, the median is robust, the mode suits categories. Range, variance and standard deviation measure spread. Correlation r lies between -1 and 1 and captures linear association only.
- Critical reading: check where the axis starts, the scale, the sample size, the source and whether the sample is representative. Truncating a y-axis exaggerates differences; 3-D pie charts distort area perception.
- Correlation is not causation: a confounding third variable or reverse causation can generate the same association.
Worked example
A school holds average marks for five classes and monthly rainfall for a year. Marks per class → bar chart, because classes are discrete categories being compared. Rainfall by month → line chart, because months form an ordered time axis and the trend matters. Heights of 200 students → histogram, because height is continuous and the shape of the distribution matters. A scatter of study hours against marks with r = 0.62 shows a moderate positive linear association, but it does not prove that studying causes higher marks.
Traps that cost marks
- A pie chart with twelve slices, or pie charts comparing separate wholes.
- A histogram used for categories, or a bar chart used for a continuous distribution.
- Joining bar tops with a line for non-temporal data, implying a trend that does not exist.
- Claiming causation from a correlation coefficient.
- Ignoring a truncated axis that visually doubles a small difference.