Data Visualization MCQs

A heavy tail decreases slowly, leaving more probability for extreme values.
This makes large events more common than under light-tailed models.

Bandwidth determines how widely each observation's kernel is spread.
It has a much larger effect on the estimate than the exact kernel shape.

Parametric estimation assumes a family such as normal or exponential.
The sample is then used to estimate a limited set of parameters.

Heteroscedasticity means conditional variability is not constant.
A funnel-shaped point cloud is a common visual sign.

Alpha sets the rejection threshold before examining the result.
Common choices include 0.05, but context should guide the value.

The k-means objective is not convex over assignments and centroids together.
Multiple random starts improve the chance of finding a better solution.

Time-series observations are indexed by time and are often dependent across periods.
Reordering them can destroy meaningful temporal structure.

Power-law relationships can appear approximately linear on logarithmic axes.
Departures from linearity indicate that a pure Pareto tail may be inadequate.

Too little smoothing follows random sample fluctuations closely.
This high-variance estimate can suggest modes that are not real.

A suitable parametric model can be efficient, concise, and easy to extrapolate.
Its assumptions must be checked rather than accepted automatically.

A smoother helps reveal trends that may be hidden by point noise.
It should be shown together with data and interpreted with uncertainty.

A Type I error is a false positive conclusion under the test framework.
Its long-run probability is controlled by the significance level when assumptions hold.