Data Analysis & Reporting Flashcards
7 cards from real AML practice questions. Tap to flip, then mark Knew It or Still Learning — missed cards come back until you master them.
Read the first 7 Data Analysis & Reporting flashcards as text
What is the purpose of a Q-Q (quantile-quantile) plot in data analysis?
Answer: Assessing whether data follows a theoretical distribution
A Q-Q plot compares the quantiles of sample data against a theoretical distribution (e.g., normal) to assess distributional fit.
In reporting model performance for an imbalanced dataset (95% negative class), which metric is most informative?
Answer: Area Under the ROC Curve (AUC-ROC)
AUC-ROC evaluates discrimination ability across all thresholds and is robust to class imbalance, unlike accuracy which can be misleadingly high.
What does a Variance Inflation Factor (VIF) greater than 10 indicate?
Answer: Severe multicollinearity in that predictor
VIF > 10 is a common threshold indicating severe multicollinearity, meaning the predictor is nearly a linear combination of others.
Which statistical test is used to determine whether two independent samples have the same mean?
Answer: Independent samples t-test
The independent samples t-test compares means of two unrelated groups to determine if they differ significantly.
A scatter plot shows a fan-shaped pattern in model residuals vs. fitted values. What does this indicate?
Answer: Heteroscedasticity
A fan-shaped residual pattern indicates heteroscedasticity, meaning residual variance increases with fitted values, violating OLS assumptions.
In exploratory data analysis (EDA), what is the primary purpose of a correlation heatmap?
Answer: Visualizing pairwise linear relationships between all features
A correlation heatmap displays pairwise Pearson (or Spearman) correlations as color-coded cells, revealing linear relationships and potential multicollinearity.
When presenting model results to a non-technical stakeholder, which reporting approach is most effective?
Answer: Translating metrics into business outcomes (e.g., revenue impact)
Non-technical stakeholders need business-contextualized metrics (e.g., 'This model saves $200K/year in fraud losses') rather than statistical abstractions.