Created from react-d3-starter template
Ebola Cases and Deaths — WHO/GAR Situation Report (Jan 06, 2016)
Dataset
File: data/ebola-cases-and-deaths-who-gar-sitrep(Jan 06, 2016).csv
Size: 1,316 bytes · Rows: 44 · Columns: 6
This is the only structured dataset in the project. It contains a snapshot of the 2014–2016 West African Ebola epidemic as reported in WHO/GAR situation reports, broken down by country and case definition.
Fields
| Column | Parsed type | Munzner class | Missing | Notes / tentative meaning |
|---|---|---|---|---|
Country |
string | Nominal (also works as Geographic identifier) | 0 | 11 distinct values. Includes Liberia2, which appears to be a second reporting line for Liberia (likely a historical report date), not a distinct country. Treat with care when aggregating. |
Case def. |
string | Nominal | 0 | 4 values: Confirmed, Probable, Suspected, All. All should equal the sum of the other three per country, though values should be verified before relying on this. |
Total cases |
string, numeric content | Quantitative | 1 | Cardinality 19; profile min 0, max 14,122. Stored as string; values are counts. One missing value. |
Cases last 21 days |
boolean-ish string | Quantitative (mostly 0) | 38 | Profile infers boolean, cardinality 1, all values 0. 38 of 44 rows missing. Indicates new cases in the 21 days before the report; at this snapshot the epidemic was nearly contained. |
Total deaths |
string, numeric content | Quantitative | 9 | Profile infers boolean due to sparse/empty values, but max 4,806 and the data show it is a death count. 9 rows missing. |
Country Report Date |
date string (e.g. 2015-12-27) |
Temporal | 37 | Only present on some rows (profile cardinality 7, years 2014–2015). Likely the date of the most recent country situation report; rows for sub-definitions often inherit it implicitly. |
Parsing assumptions
- The header is the first CSV record; cells may contain quoted embedded newlines, so a real CSV parser (d3.csv) is used.
Total casesandTotal deathsare parsed with+dand treated as quantitative; empty/invalid strings becomeNaNand are excluded from aggregations rather than coerced to 0.Cases last 21 daysis effectively all-zero/missing at this snapshot and is not visualized.
Relationships / join caveats
- There is only one dataset, so no cross-dataset joins exist. Do not join
on
Countryto any external geographic file without verifying key formatting (e.g.Liberia2). - Within the dataset, a composite key (
Country,Case def.) identifies each row. TheAllrows are aggregates of the other case definitions per country; this is assumed, not verified numerically.
Visualization
Type: Horizontal bar chart.
- Data filter: rows where
Case def.=All(one bar group per country). - Encodings: bar length (position + length) encodes
Total cases(Quantitative); a second, thinner bar in the same group encodesTotal deaths(Quantitative). Color (nominal) distinguishes cases vs deaths. Country labels on the y-axis (Nominal). - Interactions: hover tooltip shows exact values; missing death counts are shown as "No data" rather than zero.
- Rationale: a bar chart is appropriate for comparing a handful of categorical groups on two quantitative measures; sorting by total cases makes the country ranking immediately readable. Liberia is excluded from the ranking because its totals were reconstructed after data anomalies and its rows have no case-definition-level death counts for a like-for-like comparison — see the note in the tooltip-free legend.
- Accessibility: text labels, color-blind-safe pair, generous contrast.
Visualization Ideas
No project brief was provided. The ideas below are framed around the visible goal of the project: helping a general audience compare the burden of the 2014–2016 Ebola epidemic across countries, and see where the data is incomplete. All ideas load the CSV at runtime with d3.csv() and use only the existing columns. Fields are referred to by their CSV names.
Ranked Cases Lollipop Chart
- Main question: Which countries carried the largest case burden?
- Data fields:
Country,Total cases, filtered toCase def.=All. - Visual encoding: Horizontal lollipops, with position on a common axis for cases, sorted descending. Dot size is constant.
- Interaction/annotation: Hover shows the exact count. Annotate the top three.
- Implementation note: Replace bars with
lineandcircleelements on ascaleBandandscaleLinear. Sort withd3.descending.
Case Fatality Ratio Dot Plot
- Main question: What share of reported cases ended in death, per country?
- Data fields:
Total deaths/Total caseswhereCase def.=All,Country. - Visual encoding: Dot position on a 0–100% axis, one row per country. Dots are hollow if the ratio is not computable.
- Interaction/annotation: Tooltip shows the numerator and denominator. A reference line marks the pooled ratio.
- Implementation note: Compute the ratio only when both values are finite. Skip rows with
NaN.
Cases vs. Deaths Scatterplot
- Main question: How do deaths scale with cases, and which countries deviate?
- Data fields:
Total cases,Total deaths,Country. - Visual encoding: x = cases, y = deaths, point labels by country. Optionally use log scales (
scaleSymlog) so small countries stay visible. - Interaction/annotation: Draw a dashed y = x line, plus a toggle between linear and symlog axes.
- Implementation note: Use
scaleSymlogto handle zeros. Drop rows with missing deaths and list them in a footnote.
Case Definition Stacked Bars
- Main question: How is each country's total split between confirmed, probable and suspected cases?
- Data fields:
Country,Case def.(Confirmed, Probable, Suspected),Total cases. - Visual encoding: Horizontal stacked bars with a categorical color for each definition.
- Interaction/annotation: Hover a segment to highlight that definition across countries. Add a toggle for absolute counts or 100% stacking.
- Implementation note: Pivot with
d3.rollupand used3.stackwith keys in a fixed order.
Normalized Case Mix (100% Stacked)
- Main question: Which countries rely most on probable or suspected classification rather than lab confirmation?
- Data fields:
Country,Case def.,Total cases. - Visual encoding: 100% stacked horizontal bars, sorted by confirmed share.
- Interaction/annotation: Annotate countries with unusually high suspected share, such as the Liberia lines.
- Implementation note: Use
d3.stackOffsetExpand. Exclude countries whose parts sum to zero.
All vs. Sum-of-Parts Reconciliation Chart
- Main question: Does the
Allrow equal Confirmed + Probable + Suspected for each country? - Data fields:
Country,Case def.,Total cases. - Visual encoding: Diverging bar of the difference (All minus sum of parts) around a zero baseline.
- Interaction/annotation: Tooltip lists the four component values. Highlight non-zero differences in a warning color.
- Implementation note: Group with
d3.groupby country, then compute the difference. This turns the README's "assumed, not verified" caveat into something checkable.
- Main question: Does the
Data Completeness Heatmap
- Main question: Which fields are missing for which country and case definition?
- Data fields:
Country,Case def., and the presence ofTotal cases,Total deaths,Country Report Date. - Visual encoding: A grid with rows = country × case definition and columns = field. Filled vs. empty cells use two colors (nominal).
- Interaction/annotation: Hover a cell for the raw value. Add a column summary count of missing values.
- Implementation note: Treat empty strings as missing. Render with
rectcells on twoscaleBands.
Report Date Timeline
- Main question: How current is each country's reporting?
- Data fields:
Country,Country Report Date. - Visual encoding: A dot on a time axis, one row per country with a report date. Color encodes recency.
- Interaction/annotation: Annotate the snapshot date, Jan 6, 2016, as a vertical rule. Label the gap in days.
- Implementation note: Parse with
d3.timeParse('%Y-%m-%d')and usescaleTime. Only rows that carry a date are plotted.
Cases Bubble Timeline by Report Date
- Main question: Do countries with older reports also have smaller or larger totals?
- Data fields:
Country Report Date,Total cases,Country. - Visual encoding: x = report date, y = country, circle area =
Total cases(scaleSqrt). - Interaction/annotation: Tooltip shows the date and count. A note flags that the dates are not comparable across countries.
- Implementation note: Join the date from the confirmed row with
Alltotals for the sameCountry.
Treemap of Total Cases
- Main question: What is each country's share of all reported cases?
- Data fields:
Country,Total caseswhereCase def.=All. - Visual encoding: Rectangle area proportional to cases, with a sequential color for the death ratio where available.
- Interaction/annotation: Click to zoom into that country's case definitions, using the other rows as children.
- Implementation note: Build a hierarchy with
d3.hierarchyandd3.treemap. HandleLiberia2explicitly, either as a separate tile or merged with a note.
Waffle Chart of Deaths vs. Surviving Cases
- Main question: For one chosen country, how many of every 100 cases ended in death?
- Data fields:
Total cases,Total deaths,Country. - Visual encoding: A 10×10 grid of squares with death-colored vs. other squares, rounded proportionally.
- Interaction/annotation: A dropdown selects the country. A caption states the rounding.
- Implementation note: Use
Math.round(100 * deaths / cases). Disable countries without both values.
Small Multiples by Country
- Main question: What does each country's case-definition profile look like side by side?
- Data fields:
Country,Case def.,Total cases. - Visual encoding: One small bar chart per country (Confirmed, Probable, Suspected, All) with a shared y scale.
- Interaction/annotation: A toggle switches between a shared and a free y scale, so smaller countries become readable.
- Implementation note: Use a CSS grid or computed offsets and
d3.groupto split data per country.
Slope Chart: Confirmed vs. All
- Main question: How much larger is the total than the confirmed subset?
- Data fields:
Country,Total casesforCase def.=ConfirmedandAll. - Visual encoding: Two vertical axes joined by lines per country, with slope showing the gap.
- Interaction/annotation: Hover highlights a country's line and dims the others.
- Implementation note: Use two
scaleLinear(or one sharedscaleSymlog) axes. Draw lines withd3.line.
Death Reporting Gap Chart
- Main question: Where are death counts missing, and how much does that limit comparison?
- Data fields:
Country,Case def.,Total deaths,Total cases. - Visual encoding: Dumbbell or strip per country and definition. Rows with no death value appear as a hatched "No data" marker.
- Interaction/annotation: An annotation explains that Liberia has deaths only at the
Alllevel. - Implementation note: Use an SVG
patternfor hatching. Never coerce missing values to zero.
Cumulative Cases Pareto Chart
- Main question: How concentrated is the epidemic in a few countries?
- Data fields:
Country,Total cases(All). - Visual encoding: Bars sorted descending plus a cumulative-percentage line on a secondary axis.
- Interaction/annotation: Mark the 80% threshold with a reference line and label the country where it is crossed.
- Implementation note: Use
d3.cumsumon the sorted values. Normalize by the sum of finite totals.
Cases vs. Deaths Diverging Bar
- Main question: What is the gap between cases and deaths in each country, and how do the two relate in size?
- Data fields:
Country,Total cases,Total deaths. - Visual encoding: A butterfly layout with cases extending right and deaths extending left from a central axis.
- Interaction/annotation: Hover shows both values. Sort buttons switch between cases, deaths and ratio.
- Implementation note: Use two mirrored
scaleLinears sharing a domain. Re-sort with a transition.
Radial Bar Chart of Cases
- Main question: How do countries compare in a compact, circular layout?
- Data fields:
Country,Total cases(All). - Visual encoding: Radial bars with angle by country and length (
scaleRadial) by cases. - Interaction/annotation: Hover shows a tooltip. The center displays the total of all countries.
- Implementation note: Use
d3.arcwithscaleBandfor angle.scaleRadialkeeps the area perceptually accurate.
Country Detail Card with Linked Selection
- Main question: For a selected country, what are all its figures and caveats?
- Data fields: All six columns for the selected
Country. - Visual encoding: An overview bar chart on the left linked to a detail panel on the right with mini bars per
Case def., deaths and report date. - Interaction/annotation: Click a bar to update the panel. Show a "No data" badge for missing fields.
- Implementation note: Keep
selectedin React state. Both views read from the same parsed array.
Data Quality Annotation Layer
- Main question: Which numbers should viewers treat cautiously?
- Data fields:
Country(includingLiberia2),Total cases,Total deaths,Cases last 21 days. - Visual encoding: Overlay markers (a flag icon or dashed outline) on the main bar chart for rows that are missing values, are duplicate-like (
Liberia2), or have sparse fields. - Interaction/annotation: A toggle turns the caveat layer on and off. Clicking a flag opens a short explanation.
Cases last 21 daysis described as all zero/missing, so it is not plotted. - Implementation note: Compute flags from the parsed data (
isNaN, a name pattern match). Draw them as a separateglayer.
Metric Switcher Bar Chart with Animated Re-sorting
- Main question: How does the country ranking change depending on the metric?
- Data fields:
Country,Total cases,Total deaths, and the derived death ratio. - Visual encoding: One bar chart whose measure is chosen by a segmented control, with bars re-sorted on each change.
- Interaction/annotation: An animated transition (about 600 ms) shows rank changes. Countries without a value for the chosen metric are moved to the bottom with a "No data" label.
- Implementation note: Use a keyed
joinonCountrywithtransition(). Recompute the scale domain per metric.
Comments
Sign in to join the discussion.