What is a t-test commonly used to evaluate in data analysis?

Whether means differ under a stated hypothesis

A variable has a mean much larger than its median. What should an analyst suspect first?

The distribution may be right-skewed or influenced by high outliers

A company records signup plan type for visitors from three marketing channels and wants to know whether plan choice is associated with channel. What analysis is most appropriate?

A chi-squared independence test on a channel by plan table

How can learning curves help diagnose bias and variance problems?

They compare training and validation performance as data or complexity changes

What does multivariate analysis examine during exploratory data analysis?

Relationships among two or more variables

A payments company trains a fraud model, but the rare fraud class is missed because ordinary training emphasizes common legitimate transactions. What should be considered?

A class-weighted loss or sampling strategy aligned with fraud detection goals

What makes a KPI more useful for decision-making?

It is specific, measurable, aligned to a goal, time-bound, and actionable

What does compliance mean in a data and AI business context?

Following applicable laws, policies, contracts, and governance requirements

A company must assign delivery routes while minimizing fuel cost, but each driver has shift limits and each truck has capacity limits. What formulation is most appropriate?

A constrained optimization problem with cost as the objective and driver and vehicle rules as constraints

Why are convex optimization problems generally easier to reason about than nonconvex problems?

A local optimum in a convex problem is also globally optimal under the objective

CompTIA DataAI (DY0-001) Practice Test: 10 Free Questions (2026)

Exam-specific sample

Questions, answers, and option-by-option explanations

Each explanation identifies why the correct option fits and why the distractors do not.

Easy Mathematics and statistics

Question 1

What is a t-test commonly used to evaluate in data analysis?
1. Whether means differ under a stated hypothesis
2. Whether two categorical variables are independent
3. Whether a model has the highest AUC
4. Whether missing values are missing completely at random
Correct answer: A. Whether means differ under a stated hypothesis
- A. Correct. A t-test evaluates mean differences relative to variability and sample size.
- B. Incorrect. Independence between categorical variables is commonly evaluated with a chi-squared test.
- C. Incorrect. AUC evaluates binary classifier ranking performance, not mean differences.
- D. Incorrect. Missingness mechanisms require missing-data analysis, not a basic t-test.
Medium Modeling, analysis, and outcomes

Question 2

A variable has a mean much larger than its median. What should an analyst suspect first?
1. The variable must be binary
2. The dataset has no valid observations
3. The feature is certain to be causal
4. The distribution may be right-skewed or influenced by high outliers
Correct answer: D. The distribution may be right-skewed or influenced by high outliers
- A. Incorrect. Binary variables do not have to show this pattern.
- B. Incorrect. Mean and median imply observations exist.
- C. Incorrect. Summary statistics do not establish causation.
- D. Correct. Large high-end values can pull the mean above the median.
Hard Mathematics and statistics

Question 3

A company records signup plan type for visitors from three marketing channels and wants to know whether plan choice is associated with channel. What analysis is most appropriate?
1. A paired t-test on channel names
2. A chi-squared independence test on a channel by plan table
3. A one-sample t-test on total visitors
4. A regression metric such as RMSE
Correct answer: B. A chi-squared independence test on a channel by plan table
- A. Incorrect. Channel names are categories, not paired continuous measurements.
- B. Correct. Both channel and plan type are categorical, so a contingency-table association test fits.
- C. Incorrect. A one-sample t-test would compare a mean to a benchmark, not test association.
- D. Incorrect. RMSE evaluates numeric prediction error and does not test categorical association.
Medium Machine learning

Question 4

How can learning curves help diagnose bias and variance problems?
1. They replace labels with predicted scores
2. They show whether a database backup succeeded
3. They compare training and validation performance as data or complexity changes
4. They convert regression into classification
Correct answer: C. They compare training and validation performance as data or complexity changes
- A. Incorrect. Learning curves do not replace labels.
- B. Incorrect. Backup status is not model diagnostics.
- C. Correct. Learning curves reveal underfitting, overfitting, and data-volume effects.
- D. Incorrect. Learning curves do not change task type.
Easy Modeling, analysis, and outcomes

Question 5

What does multivariate analysis examine during exploratory data analysis?
1. Relationships among two or more variables
2. One variable in isolation only
3. The access permissions on a table
4. The exact deployment region for a model
Correct answer: A. Relationships among two or more variables
- A. Correct. Multivariate analysis explores how variables relate to each other.
- B. Incorrect. One-variable analysis is univariate analysis.
- C. Incorrect. Permissions are governance or security metadata.
- D. Incorrect. Deployment region is infrastructure context.
Hard Machine learning

Question 6

A payments company trains a fraud model, but the rare fraud class is missed because ordinary training emphasizes common legitimate transactions. What should be considered?
1. Using transaction IDs as the loss function
2. Optimizing only overall accuracy on the training set
3. A class-weighted loss or sampling strategy aligned with fraud detection goals
4. Dropping rare fraud labels from the dataset
Correct answer: C. A class-weighted loss or sampling strategy aligned with fraud detection goals
- A. Incorrect. Transaction IDs are identifiers, not losses.
- B. Incorrect. Overall training accuracy can ignore rare fraud.
- C. Correct. Rare positive classes often need loss or sampling adjustments.
- D. Incorrect. Dropping fraud labels removes the signal.
Medium Operations and processes

Question 7

What makes a KPI more useful for decision-making?
1. It is difficult for stakeholders to interpret
2. It is specific, measurable, aligned to a goal, time-bound, and actionable
3. It changes definition whenever results are unfavorable
4. It is calculated from whichever source is easiest that day
Correct answer: B. It is specific, measurable, aligned to a goal, time-bound, and actionable
- A. Incorrect. Poor interpretability reduces usefulness.
- B. Correct. Useful KPIs are clearly defined and tied to decisions.
- C. Incorrect. Changing definitions breaks comparability.
- D. Incorrect. Inconsistent sources create unreliable metrics.
Easy Operations and processes

Question 8

What does compliance mean in a data and AI business context?
1. Maximizing model accuracy regardless of data-use restrictions
2. Selecting the fastest algorithm for every workflow
3. Publishing model outputs before stakeholder review
4. Following applicable laws, policies, contracts, and governance requirements
Correct answer: D. Following applicable laws, policies, contracts, and governance requirements
- A. Incorrect. Accuracy does not override legal or policy duties.
- B. Incorrect. Speed is not the definition of compliance.
- C. Incorrect. Unreviewed publishing can create compliance risk.
- D. Correct. Compliance is about meeting required obligations and controls.
Hard Specialized applications of data science

Question 9

A company must assign delivery routes while minimizing fuel cost, but each driver has shift limits and each truck has capacity limits. What formulation is most appropriate?
1. An unconstrained text-classification problem using route descriptions as labels
2. A clustering problem that ignores truck capacity during assignment
3. A constrained optimization problem with cost as the objective and driver and vehicle rules as constraints
4. A regression model that predicts route cost independently and assigns all routes to the cheapest truck without checking limits
Correct answer: C. A constrained optimization problem with cost as the objective and driver and vehicle rules as constraints
- A. Incorrect. Route assignment with capacity and shift limits is not a text-classification task.
- B. Incorrect. Ignoring capacity can produce infeasible delivery plans.
- C. Correct. The decision must optimize cost while respecting operational constraints.
- D. Incorrect. Cost prediction alone does not solve the assignment; the plan must optimize cost while jointly respecting driver shifts and truck capacities.
Medium Specialized applications of data science

Question 10

Why are convex optimization problems generally easier to reason about than nonconvex problems?
1. A local optimum in a convex problem is also globally optimal under the objective
2. Convex problems do not require explicit objective functions or feasibility constraints
3. Nonconvex problems cannot be used in machine learning
4. Convexity means the dataset has no missing values
Correct answer: A. A local optimum in a convex problem is also globally optimal under the objective
- A. Correct. Convex structure gives stronger optimality reasoning under the objective.
- B. Incorrect. Convex optimization still needs an objective.
- C. Incorrect. Numerous machine learning models involve nonconvex optimization.
- D. Incorrect. Convexity is about the objective landscape, not missing data.

CompTIA DataAI (DY0-001) Practice Test: 10 Free Questions

Questions, answers, and option-by-option explanations

Question 1

Question 2

Question 3

Question 4

Question 5

Question 6

Question 7

Question 8

Question 9

Question 10

Continue with the full exam-prep cluster

How can I help?