ISYS3438 Chap.8 Temporal, Hierarchical and Text Data
Temporal, Hierarchical and Text Data
Time, hierarchy and text contain internal structure that ordinary category-versus-measure selection can erase. This chapter distinguishes trends and cycles, parent-child containment and words in context. It compares lines, area charts and calendar heatmaps for temporal work, tree diagrams and treemaps for hierarchies, and ranked term views or word clouds for text.
Tableau parameters are treated as decision controls whose value lies in the calculation, filter or reference question they change.
What this chapter covers
- 01
time order and cycles
- 02
calendar heatmaps
- 03
tree diagrams
- 04
treemaps
- 05
text tokenisation
- 06
parameters
Worked example · free
Coordinate returns across three structures
- +1Use a time trend plus calendar heatmap for sequence and weekly seasonality.
- +1Use a hierarchy view for containment and a sorted bar for precise share comparison.
- +1Document token rules and retain contextual quotations for text interpretation.
- +1Let a visible parameter switch between count and rate with the denominator printed.
Key terms
- Calendar heatmap
- A calendar heatmap maps a value to calendar cells to reveal recurring temporal patterns.
- Treemap
- A treemap uses nested areas to show hierarchical containment and part-to-whole magnitude.
- Tokenisation
- Tokenisation splits text into analytical units under an explicit rule.
- Parameter
- A parameter is a user-controlled value that can alter a calculation, filter or reference line.
Temporal, Hierarchical and Text Data FAQ
When is a calendar heatmap useful?
It is useful when recurring weekday or calendar structure matters. Pair it with a line when the audience also needs the uninterrupted sequence and long-term trend.
Why is a treemap hard to compare?
Area comparison is less accurate than common-position length, especially for similar rectangles. Treemaps are effective for containment and overview, while a sorted bar or table supports precise ranking.
Can a word cloud show sentiment?
Not by itself. Word size usually represents frequency, and token choices remove context such as negation or speaker. Use it for orientation, then inspect contextual text or a documented coding method.
Assessment move
Build a small reference table with data form, preserved structure, useful tasks and common failure. For time, test missing periods and interval consistency. For hierarchy, check parent-child totals. For text, save token rules and examples of lost context. For parameters, write the decision that changes when the control moves. Match the display to the structure that must be preserved.
Time has order and often regular intervals; hierarchy has parent–child containment; text has tokens, context and language ambiguity. Write the analytical task before selecting a specialised chart. A familiar bar chart may outperform a decorative form when the decision is a simple comparison, while a calendar heatmap, tree or term matrix earns its place only when its structural pattern matters.
For temporal data, verify time zone, missing periods, interval length and whether values are stocks, flows or cumulative totals. Use a line for continuous sequence, bars for discrete periods and a calendar heatmap for recurring weekday or seasonal patterns. Do not connect gaps as though observations existed. Mark external events with sourced annotations and distinguish temporal alignment from causal effect.
Compare absolute change, percentage change and an appropriate baseline before choosing the claim. For hierarchies, validate that children belong to one parent or document any many-to-many relationship. Trees reveal paths and levels; treemaps provide compact part-to-whole overview but weaken exact comparison, especially among differently shaped rectangles.
Keep depth manageable and offer a common-scale detail view when ranking matters. Check whether totals reconcile across levels and whether filtering removes a parent while leaving its descendants misleadingly prominent. For text, preserve enough context to interpret frequency. A word cloud emphasises common tokens through area but makes close comparisons difficult and can elevate boilerplate.
Clean case, punctuation and stop words transparently, retain negation where it changes meaning and compare term rates across groups rather than raw counts alone. Validate automated sentiment or topic labels against a sample of original passages. In Tableau, document parameter values and ensure the audience can see the current state.
Finish by pairing any specialised overview with a table, excerpt or trend that supports verification. Use a structure-preservation checklist before interpretation. For time, confirm order, interval and missing periods; for hierarchy, confirm parentage, totals and depth; for text, confirm cleaning, context and validation. Rebuild one important result in a simpler companion view such as a table, line or sorted bar chart.
If the specialised form and companion disagree, inspect aggregation, filtering and transformation before writing the claim. Describe why the specialised display adds useful structure beyond the simpler view. This comparison strengthens a report because it demonstrates that the chosen form was earned by the question, not selected for novelty, and that its most important pattern can still be independently verified.
For revision, take one dataset and write three separate questions that genuinely require time, hierarchy and text. Build only the minimum view for each, then explain why combining them could obscure their distinct structures. This exercise develops restraint and helps you recognise when coordinated views are preferable to one overloaded specialised display.