NCA-ADS - Descriptive Analysis and Visualization (13% of the exam) - Section 4.4
Interpret patterns and trends.
Interpret patterns and trends in data by identifying seasonality, monotonic trends, clusters, and anomalies from visualisations and summary outputs. Distinguish between correlation and causation when drawing conclusions from observed relationships.
Practice question for this objective
lock_openFree sampleDescriptive Analysis and Visualizationeasy
While exploring a cuDF DataFrame of bike-share trips, an analyst wants to understand how trip duration relates to the distance travelled, with one point per trip across 100,000 trips. Both columns are continuous numeric variables. Which chart type best reveals the relationship and any correlation between these two variables?
AA histogram of trip durations with a separate histogram of trip distances placed beside it.
BA bar chart with one bar for the total duration and one bar for the total distance.
CA box plot of trip duration with the trip distance encoded as the whisker length.
DA scatter plot with distance on one axis and duration on the other, one marker per trip.check_circle Correct
A scatter plot is the appropriate chart for examining the relationship and correlation between two continuous numeric variables. A scatter plot places each observation at the coordinates of its two variable values, so the pattern of the resulting point cloud reveals whether the variables rise together, fall together, or show no relationship, which single-variable charts cannot do.
Why A is wrong: Tempting because histograms describe each variable's distribution, but two side-by-side histograms show the variables in isolation and cannot expose how they move together.
Why B is wrong: Tempting because bar charts are simple to read, but two summed bars collapse 100,000 trips into two totals and convey nothing about the per-trip relationship.
Why C is wrong: Tempting because box plots summarise spread, but a box plot displays the distribution of one variable and its whisker length encodes that variable's range, not a second variable's values.
Why D is correct: Correct. A scatter plot maps two continuous variables onto perpendicular axes so the cloud of points reveals the direction, strength, and shape of any relationship between distance and duration.
Exam traps in Descriptive Analysis and Visualization
Answers that look right on this material and are not. Each one is a distractor from a different question in the NCA-ADS bank for this domain.
A stacked bar chart, with one bar per feature divided into 15 coloured segments representing its correlation with each feature.
errorWhy it is wrong: A stacked bar chart shows composition within categories and is tempting because it uses colour segments, but stacking signed correlation values is misleading and comparing 15 segments per bar is far harder than reading a coloured grid.
Each additional unit of advertising spend causes weekly sales to rise by 0.85 units, holding everything else constant.
errorWhy it is wrong: This conflates correlation with causation and with a regression slope. The coefficient measures association strength, not a causal effect or a unit-for-unit rate of change.
The near-zero Pearson value proves the two variables are independent, so the inverted-U shape in the scatter plot must be a plotting artefact.
errorWhy it is wrong: Tempting because zero correlation is loosely equated with no relationship, but Pearson measures only linear association, and a clear deterministic curve means the variables are strongly related, not independent.