DSE Exploratory Data Analysis Techniques 5 — Questions and Answers
Question 1: Which plot is most suitable for visualizing changes in a continuous variable over time?
- Box plot
- Line chart (Correct answer)
- Pie chart
- Histogram
Correct answer: Line chart
A line chart connects sequential data points in chronological order, making trends, cycles, and anomalies in time-series data easy to see.
Question 2: What is the purpose of a pivot table in EDA?
- To impute missing values with group medians
- To summarize data by aggregating values across two or more categorical dimensions (Correct answer)
- To detect multicollinearity between predictors
- To transform skewed features into a normal distribution
Correct answer: To summarize data by aggregating values across two or more categorical dimensions
A pivot table reorganizes data into a matrix where rows and columns represent categorical variables and cells contain aggregated statistics like counts or means.
Question 3: A dataset has a bimodal distribution. What does this likely indicate?
- The data was collected with measurement error
- The dataset may contain two distinct subpopulations or groups (Correct answer)
- The mean and median are equal
- The variance is unusually high
Correct answer: The dataset may contain two distinct subpopulations or groups
Two peaks (modes) in a distribution often suggest the presence of two distinct subgroups with different characteristic values mixed into a single dataset.
Question 4: Which of the following is a key step in univariate EDA for a continuous variable?
- Computing a chi-square test between two categories
- Examining the distribution shape, central tendency, spread, and outliers (Correct answer)
- Building a decision tree model on the variable
- Calculating pairwise correlations with all other variables
Correct answer: Examining the distribution shape, central tendency, spread, and outliers
Univariate EDA for a continuous variable involves summarizing and visualizing its distribution shape, mean/median, standard deviation, and extreme values.
Question 5: What does Spearman's rank correlation measure that Pearson's correlation does not?
- The linear relationship between two variables
- Monotonic relationships, including nonlinear ones, between two variables (Correct answer)
- The covariance between two variables scaled by their variances
- The proportion of variance explained by a linear fit
Correct answer: Monotonic relationships, including nonlinear ones, between two variables
Spearman's correlation computes Pearson's r on the ranks of the data, capturing any monotonic relationship (not just linear) and being robust to outliers.
Question 6: In EDA, what does 'cardinality' of a categorical variable refer to?
- The proportion of missing values in the column
- The number of unique distinct values the variable can take (Correct answer)
- The frequency of the most common category
- The correlation between the variable and the target
Correct answer: The number of unique distinct values the variable can take
Cardinality is the count of distinct categories in a categorical variable; high-cardinality variables (e.g., user IDs) may need special encoding strategies.
Question 7: Which technique helps identify the most important variables early in EDA without building a full model?
- Replacing missing values with the column mean
- Computing correlation with the target variable or using mutual information scores (Correct answer)
- Normalizing all features to the same scale
- Splitting the data into training and test sets
Correct answer: Computing correlation with the target variable or using mutual information scores
Correlations and mutual information scores quantify how much each feature relates to the target, providing a fast model-free signal for feature relevance.
DSE Exploratory Data Analysis Techniques 5
Which plot is most suitable for visualizing changes in a continuous variable over time?