WGU D492 Data Analytics - Applications | proctored CompTIA
Data+ DA0-002 certification exam study guide | Questions and
Answers | 2026 Update | 100% Correct. - 234 Questions and
Answers Already Graded A+ Premium Exam Tested And
Verified
Subject Area Data Analytics - Applications (CompTIA Data+ aligned)
Description This exam covers core data analytics concepts including data governance, data
quality, ETL/ELT processes, statistical analysis, data mining, visualization, and
database querying. It is designed to assess advanced understanding of data
analytics applications at a level equivalent to the CompTIA Data+ DA0-002
certification.
Expected Grade A+
Total Questions 234
Duration 3 hours
Learning Outcomes 1. Apply data governance and quality frameworks to real-world scenarios
2. Evaluate and select appropriate data analysis methods and statistical tests
3. Design and interpret data visualizations and reports
4. Implement ETL processes and database queries for data preparation
5. Identify and mitigate bias and errors in data sampling and analysis
Accreditation US university standards - WGU D492 Data Analytics - Applications course
Page 1
,1. A dataset containing customer addresses has 5% missing values in the 'state'
column, 2% missing in 'zip code', and 1% missing in 'city'. Additionally, 3% of
records have a city that does not match the state. Which data quality
dimension is most severely compromised?
Answer: Completeness
Completeness is the extent to which data is not missing. With 5% missing state
and 2% missing zip, the overall missing rate is significant, making completeness
the worst dimension. Consistency (city-state mismatch 3%) is less severe.
Accuracy would require validation against a trusted source, not just internal
checks. Timeliness is not addressed.
2. A data engineer needs to load raw transaction data into a data warehouse
for real-time dashboards. The transformations required are complex and
involve cleansing and joining multiple streaming sources. Which approach is
most appropriate?
Answer: ETL (Extract, Transform, Load) because transformations can be
applied before loading, ensuring clean data in the warehouse.
For real-time dashboards, low latency and clean data are critical. ETL
transforms data before loading, ensuring the warehouse contains clean,
ready-to-use data, supporting real-time queries without on-the-fly transformation
overhead. ELT is better for ad-hoc analysis but adds latency for transformations.
Staging and view-based approaches do not meet real-time requirements
efficiently.
3. In a clinical trial, researchers test a new drug against a placebo on 20
different outcome measures. They report a p-value of 0.03 for one outcome as
statistically significant. What is the primary concern with this conclusion?
Answer: Multiple comparisons increase the probability of a Type I error.
Testing 20 outcomes inflates the familywise error rate; with =0.05, there is a high
chance of at least one false positive. A p-value of 0.03 is not significant after
correcting for multiple comparisons (e.g., Bonferroni adjusts threshold to
0.0025). The p-value itself is below 0.05, so A is false. Power concerns may exist
but are secondary; effect size is important but not the primary issue.
Page 2
,4. A linear regression model predicting house prices has residuals that increase
in variance as the predicted price increases. Which regression assumption is
violated?
Answer: Homoscedasticity
Homoscedasticity requires constant variance of residuals across all levels of
predicted values. Increasing variance indicates heteroscedasticity. Linearity
(relation between predictors and outcome) might still hold. Normality violation
would appear in a Q-Q plot, not as a funnel shape. Independence would be
violated in time series or clustered data.
5. A decision tree trained on a dataset with 50 features and 10,000 records
achieves 99% accuracy on training data but only 75% on test data. Which
action is most likely to improve test set performance without significant loss of
interpretability?
Answer: Apply pruning (e.g., cost-complexity pruning) to reduce overfitting.
The large gap indicates overfitting. Pruning reduces tree complexity, improving
generalization. Increasing depth (A) worsens overfitting. Adding more features
(C) may increase noise. A random forest (D) can improve accuracy but reduces
interpretability, which is specified as important.
6. In K-means clustering, the elbow method shows an elbow at k=4, while the
silhouette score is highest for k=6. The dataset has known natural clusters that
are non-spherical. Which evaluation should you trust and why?
Answer: Neither is reliable because K-means assumes spherical clusters of
equal size; use a different algorithm like DBSCAN.
K-means performs poorly on non-spherical clusters; both elbow and silhouette
are based on distance measures that assume globular clusters. Therefore, both
metrics may be misleading. DBSCAN or spectral clustering would be more
appropriate. The elbow method (A) and silhouette (B) both suffer from the same
assumption. Silhouette is less sensitive to noise than sometimes thought but still
flawed for non-spherical shapes.
Page 3
, 7. A data analyst wants to visualize survey responses on a 5-point Likert scale
(Strongly Disagree to Strongly Agree) for multiple questions. The goal is to
highlight overall agreement versus disagreement patterns across questions.
Which chart type is most effective?
Answer: Diverging bar chart with neutral in the center and
agreement/disagreement diverging.
A diverging bar chart positions neutral responses at a center line, with agreement
extending to one side and disagreement to the other, making it easy to compare
net sentiment across questions. Stacked bars (A) require mental arithmetic to
compare net agreement. Pie charts (C) are poor for comparing multiple
questions. Line chart (D) loses distribution information and assumes interval
scale, which may be inappropriate for Likert data.
8. Given tables: Customers (CustomerID, Name, City), Orders (OrderID,
CustomerID, OrderDate, Amount). Write a query to find customers who
placed more than 3 orders in 2023, but with total amount less than $500.
Which SQL clause correctly implements the condition on total amount?
Answer: HAVING SUM(Amount) < 500
Once aggregation with GROUP BY is applied, conditions on aggregate results
must use HAVING, not WHERE. HAVING SUM(Amount) < 500 is correct
syntax. B uses WHERE with aggregate function, which is invalid. C uses alias
'totalAmount' but alias cannot be used directly in HAVING unless defined in the
same query level; best practice is to use the expression. D uses WHERE with
alias, also invalid.
9. A survey on employee satisfaction is mailed to 10,000 employees; 2,000
respond. The responding group has a higher proportion of part-time employees
compared to the overall population. Which type of bias is most likely present?
Answer: Selection bias due to non-response
Non-response bias occurs when respondents differ systematically from
non-respondents. Here, the over-representation of part-time employees indicates
that the sample is not representative of the population. Measurement bias (B)
relates to the survey instrument, not differential response. Confirmation bias (C)
is an analyst error. Recall bias (D) is about memory, not participation.
Page 4
Data+ DA0-002 certification exam study guide | Questions and
Answers | 2026 Update | 100% Correct. - 234 Questions and
Answers Already Graded A+ Premium Exam Tested And
Verified
Subject Area Data Analytics - Applications (CompTIA Data+ aligned)
Description This exam covers core data analytics concepts including data governance, data
quality, ETL/ELT processes, statistical analysis, data mining, visualization, and
database querying. It is designed to assess advanced understanding of data
analytics applications at a level equivalent to the CompTIA Data+ DA0-002
certification.
Expected Grade A+
Total Questions 234
Duration 3 hours
Learning Outcomes 1. Apply data governance and quality frameworks to real-world scenarios
2. Evaluate and select appropriate data analysis methods and statistical tests
3. Design and interpret data visualizations and reports
4. Implement ETL processes and database queries for data preparation
5. Identify and mitigate bias and errors in data sampling and analysis
Accreditation US university standards - WGU D492 Data Analytics - Applications course
Page 1
,1. A dataset containing customer addresses has 5% missing values in the 'state'
column, 2% missing in 'zip code', and 1% missing in 'city'. Additionally, 3% of
records have a city that does not match the state. Which data quality
dimension is most severely compromised?
Answer: Completeness
Completeness is the extent to which data is not missing. With 5% missing state
and 2% missing zip, the overall missing rate is significant, making completeness
the worst dimension. Consistency (city-state mismatch 3%) is less severe.
Accuracy would require validation against a trusted source, not just internal
checks. Timeliness is not addressed.
2. A data engineer needs to load raw transaction data into a data warehouse
for real-time dashboards. The transformations required are complex and
involve cleansing and joining multiple streaming sources. Which approach is
most appropriate?
Answer: ETL (Extract, Transform, Load) because transformations can be
applied before loading, ensuring clean data in the warehouse.
For real-time dashboards, low latency and clean data are critical. ETL
transforms data before loading, ensuring the warehouse contains clean,
ready-to-use data, supporting real-time queries without on-the-fly transformation
overhead. ELT is better for ad-hoc analysis but adds latency for transformations.
Staging and view-based approaches do not meet real-time requirements
efficiently.
3. In a clinical trial, researchers test a new drug against a placebo on 20
different outcome measures. They report a p-value of 0.03 for one outcome as
statistically significant. What is the primary concern with this conclusion?
Answer: Multiple comparisons increase the probability of a Type I error.
Testing 20 outcomes inflates the familywise error rate; with =0.05, there is a high
chance of at least one false positive. A p-value of 0.03 is not significant after
correcting for multiple comparisons (e.g., Bonferroni adjusts threshold to
0.0025). The p-value itself is below 0.05, so A is false. Power concerns may exist
but are secondary; effect size is important but not the primary issue.
Page 2
,4. A linear regression model predicting house prices has residuals that increase
in variance as the predicted price increases. Which regression assumption is
violated?
Answer: Homoscedasticity
Homoscedasticity requires constant variance of residuals across all levels of
predicted values. Increasing variance indicates heteroscedasticity. Linearity
(relation between predictors and outcome) might still hold. Normality violation
would appear in a Q-Q plot, not as a funnel shape. Independence would be
violated in time series or clustered data.
5. A decision tree trained on a dataset with 50 features and 10,000 records
achieves 99% accuracy on training data but only 75% on test data. Which
action is most likely to improve test set performance without significant loss of
interpretability?
Answer: Apply pruning (e.g., cost-complexity pruning) to reduce overfitting.
The large gap indicates overfitting. Pruning reduces tree complexity, improving
generalization. Increasing depth (A) worsens overfitting. Adding more features
(C) may increase noise. A random forest (D) can improve accuracy but reduces
interpretability, which is specified as important.
6. In K-means clustering, the elbow method shows an elbow at k=4, while the
silhouette score is highest for k=6. The dataset has known natural clusters that
are non-spherical. Which evaluation should you trust and why?
Answer: Neither is reliable because K-means assumes spherical clusters of
equal size; use a different algorithm like DBSCAN.
K-means performs poorly on non-spherical clusters; both elbow and silhouette
are based on distance measures that assume globular clusters. Therefore, both
metrics may be misleading. DBSCAN or spectral clustering would be more
appropriate. The elbow method (A) and silhouette (B) both suffer from the same
assumption. Silhouette is less sensitive to noise than sometimes thought but still
flawed for non-spherical shapes.
Page 3
, 7. A data analyst wants to visualize survey responses on a 5-point Likert scale
(Strongly Disagree to Strongly Agree) for multiple questions. The goal is to
highlight overall agreement versus disagreement patterns across questions.
Which chart type is most effective?
Answer: Diverging bar chart with neutral in the center and
agreement/disagreement diverging.
A diverging bar chart positions neutral responses at a center line, with agreement
extending to one side and disagreement to the other, making it easy to compare
net sentiment across questions. Stacked bars (A) require mental arithmetic to
compare net agreement. Pie charts (C) are poor for comparing multiple
questions. Line chart (D) loses distribution information and assumes interval
scale, which may be inappropriate for Likert data.
8. Given tables: Customers (CustomerID, Name, City), Orders (OrderID,
CustomerID, OrderDate, Amount). Write a query to find customers who
placed more than 3 orders in 2023, but with total amount less than $500.
Which SQL clause correctly implements the condition on total amount?
Answer: HAVING SUM(Amount) < 500
Once aggregation with GROUP BY is applied, conditions on aggregate results
must use HAVING, not WHERE. HAVING SUM(Amount) < 500 is correct
syntax. B uses WHERE with aggregate function, which is invalid. C uses alias
'totalAmount' but alias cannot be used directly in HAVING unless defined in the
same query level; best practice is to use the expression. D uses WHERE with
alias, also invalid.
9. A survey on employee satisfaction is mailed to 10,000 employees; 2,000
respond. The responding group has a higher proportion of part-time employees
compared to the overall population. Which type of bias is most likely present?
Answer: Selection bias due to non-response
Non-response bias occurs when respondents differ systematically from
non-respondents. Here, the over-representation of part-time employees indicates
that the sample is not representative of the population. Measurement bias (B)
relates to the survey instrument, not differential response. Confirmation bias (C)
is an analyst error. Recall bias (D) is about memory, not participation.
Page 4
