Support pairwise relationship queries with explicit attribute selection

Let viewers choose two attributes and inspect their relationship over the current set of cases. For example, support numeric associations, category-to-attribute coincidences, and trends where time is one of the selected variables.

Low-level correlation questions are usually about a specified pair of attributes over a specified set of cases. That framing keeps the task concrete, including trend questions where one of the attributes is time.

Mechanism: Explicit pair selection turns a vague search for “something related” into a concrete relationship query that the visualization can answer directly.

Evidence: Correlate is defined as determining the relationship between two attributes over a given set of data cases, the taxonomy observes that users frequently asked for both numeric and non-numeric correlations, and it interprets trend questions as correlations with temporal variables (Amar et al., 2005).

• User Goal: Determine whether two chosen attributes relate in a useful way.
• Task: Correlate a specified attribute pair over a set of cases.
• Data: Tabular data with at least two attributes, including temporal attributes when the question is about trend.
• Chart Setting: An information visualization system that can operate on the current subset of cases.
• Audience: Analysts asking concrete pairwise relationship questions.
• Success Criterion: The system supports direct inspection of the chosen pair, including time paired with another variable when needed.

Break it when: The question is open-ended about any related variable or relies on uncertain criteria. Why: Those are higher-level exploratory questions that the primitive correlate task does not directly cover.

Sacrifice: The user must specify the two attributes of interest before the system can answer the query.
Risk: Open-ended questions such as “what relates to X?” remain unanswered by this low-level control.
Mitigation: Use the pairwise control for specified questions and keep broader relationship discovery as a separate workflow.

Mistake: Treating open-ended relationship hunting as the same task as a specified pairwise correlation. Why it fails: The low-level task requires a given set of cases and two chosen attributes.

Failure Sign: Users cannot directly specify two attributes and inspect their relation on the current subset.
Quick Check: Choose two attributes and ask for their relationship on a selected set of cases.
Stronger Test: Repeat the check with time as one of the two attributes to confirm support for trend questions.

• Add explicit two-attribute selection for relationship queries.
• Let the pairwise relationship operation run on filtered or otherwise selected subsets.
• Include temporal attributes in the same pairwise relationship control.