MS-DS Master of Data science FREE Master of Data Science Research & Data Analysis Questions and Answers 2 — Questions and Answers
Question 1: A data scientist wants to determine whether a new recommendation algorithm increases user engagement compared to the existing one. Which experimental design is most appropriate?
- A/B testing with random user assignment (Correct answer)
- Observational cohort study
- Case-control study
- Time-series analysis without a control group
Correct answer: A/B testing with random user assignment
A/B testing with random assignment allows causal inference by controlling for confounding variables between the treatment and control groups.
Question 2: When performing exploratory data analysis on a dataset with 50 features, which technique is most useful for identifying multicollinearity before building a regression model?
- Variance Inflation Factor (VIF) analysis (Correct answer)
- Chi-square test of independence
- Kolmogorov-Smirnov test
- Mann-Whitney U test
Correct answer: Variance Inflation Factor (VIF) analysis
VIF quantifies how much the variance of a regression coefficient is inflated due to multicollinearity among predictors.
Question 3: A researcher notices that their linear regression model has residuals that fan out as predicted values increase. What assumption is being violated?
- Normality of residuals
- Homoscedasticity (Correct answer)
- Independence of observations
- Linearity
Correct answer: Homoscedasticity
Residuals that fan out indicate heteroscedasticity, meaning the variance of errors is not constant across predicted values.
Question 4: In a master's thesis analyzing healthcare data, which method best handles missing data when the missingness mechanism is Missing at Random (MAR)?
- Listwise deletion
- Mean imputation
- Multiple imputation (Correct answer)
- Replacing with zero
Correct answer: Multiple imputation
Multiple imputation generates several plausible datasets, accounts for uncertainty in the missing values, and produces valid statistical inferences under MAR.
Question 5: A data science graduate student is presenting research findings and reports a p-value of 0.03 with a confidence interval of [-0.5, 12.3] for the treatment effect. What is the most valid interpretation?
- The treatment definitely has a positive effect
- The result is statistically significant but the wide interval suggests low practical precision (Correct answer)
- The null hypothesis is true with 97% probability
- The treatment effect is exactly 5.9
Correct answer: The result is statistically significant but the wide interval suggests low practical precision
While the p-value indicates statistical significance at the 0.05 level, the wide confidence interval spanning near-zero to 12.3 indicates considerable uncertainty about the true effect size.
Question 6: Which cross-validation strategy is most appropriate when working with time-series data in a predictive modeling research project?
- Standard k-fold cross-validation
- Leave-one-out cross-validation
- Time-series split (expanding window) cross-validation (Correct answer)
- Stratified k-fold cross-validation
Correct answer: Time-series split (expanding window) cross-validation
Time-series split preserves the temporal ordering of observations, ensuring the model is always trained on past data and tested on future data.
