WGU D600 Exam 3: Principal Component
Analysis (PCA) | Complete Regression
Report, Python Modeling & Results |
2026/2027 {JUST RELEASED} Updated
Practice Question Bank & Answer Rationale
THIS PACK CONTAINS
50 exam-style multiple-choice questions. Each question includes the correct answer and a complete rationale
for focused revision.
COVERAGE
This exam assesses graduate-level mastery of principal component analysis (PCA) theory, dimensionality
reduction, regression modeling, and Python implementation for a complete analytics report. It covers variance
explained, component selection, standardization, multicollinearity resolution, model diagnostics, and results
communication aligned with WGU D600 Exam 3 (2026/2027).
STUDY GUIDE
Recommended duration: 3 hours. Passing target: 80%.
HIGHP - Page 1 of 18
,1. In a PCA-based regression for home prices, which variable set is most appropriate for
PCA?
[ ] A. SquareFootage, NumBedrooms, BackyardSpace, CrimeRate, SchoolRating, AgeOfHome,
DistanceToCityCenter, EmploymentRate, PropertyTaxRate, RenovationQuality, TransportAccess,
PreviousSalePrice
[ ] B. SquareFootage, NumBedrooms, BackyardSpace, CrimeRate, SchoolRating, AgeOfHome,
DistanceToCityCenter, EmploymentRate, PropertyTaxRate, RenovationQuality, TransportAccess
[ ] C. SquareFootage, NumBedrooms, BackyardSpace, CrimeRate, SchoolRating, AgeOfHome,
DistanceToCityCenter, EmploymentRate, PropertyTaxRate, RenovationQuality, TransportAccess,
PreviousSalePrice, ZipCode
[ ] D. SquareFootage, NumBedrooms, BackyardSpace, CrimeRate, SchoolRating, AgeOfHome,
DistanceToCityCenter, EmploymentRate, PropertyTaxRate, RenovationQuality, TransportAccess,
PreviousSalePrice, SaleDate
CORRECT: B. SquareFootage, NumBedrooms, BackyardSpace, CrimeRate, SchoolRating,
AgeOfHome, DistanceToCityCenter, EmploymentRate, PropertyTaxRate, RenovationQuality,
TransportAccess
PCA should be applied to continuous explanatory variables only; PreviousSalePrice is likely the target or a
leaky variable, and ZipCode/SaleDate are non-continuous. Option B correctly excludes PreviousSalePrice
and non-continuous variables, while A includes the target, and C/D add categorical or date variables
inappropriate for PCA.
2. After standardizing 12 housing variables and running PCA, the first four components
explain 85% of variance. What is the most defensible next step for regression?
[ ] A. Retain all 12 original variables to preserve interpretability.
[ ] B. Use the first four principal components as predictors in a linear regression.
[ ] C. Discard PCA and use ridge regression instead.
[ ] D. Retain components until 99% variance is explained.
CORRECT: B. Use the first four principal components as predictors in a linear regression.
Retaining components that cumulatively explain a substantial portion (e.g., 85%) is a common and defensible
rule for dimensionality reduction. Using all original variables defeats the purpose of PCA, switching to ridge
ignores the PCA goal, and requiring 99% may retain too many components and overfit.
3. Which statement best justifies standardizing variables before PCA in the housing price
analysis?
[ ] A. Standardization ensures all variables have equal weight regardless of scale.
[ ] B. Standardization increases the variance explained by the first component.
[ ] C. Standardization removes multicollinearity among variables.
[ ] D. Standardization makes the data normally distributed.
HIGHP - Page 2 of 18
, CORRECT: A. Standardization ensures all variables have equal weight regardless of scale.
PCA is sensitive to scale; standardizing to mean 0 and SD 1 gives each variable equal influence. It does not
inherently increase variance explained, remove multicollinearity (PCA does that), or normalize distributions.
4. In the regression report, the PCA-transformed predictors are orthogonal. What does this
imply about the regression coefficients?
[ ] A. Coefficients are unaffected by multicollinearity and are stable.
[ ] B. Coefficients are biased due to omitted variables.
[ ] C. Coefficients cannot be interpreted as original variable effects.
[ ] D. Coefficients are all statistically insignificant.
CORRECT: A. Coefficients are unaffected by multicollinearity and are stable.
Orthogonal components eliminate multicollinearity, yielding stable coefficient estimates. However, they are
not directly interpretable as original variable effects (so C is partially true but not the best answer), and they
are not biased or necessarily insignificant.
5. A PCA regression yields R² = 0.78 on training data but 0.45 on test data. Which action
best addresses this?
[ ] A. Add more principal components to the model.
[ ] B. Reduce the number of principal components or apply regularization.
[ ] C. Remove the standardization step.
[ ] D. Use all original variables instead of components.
CORRECT: B. Reduce the number of principal components or apply regularization.
The large gap indicates overfitting; reducing components or regularizing improves generalization. Adding
components worsens overfitting, removing standardization is inappropriate, and using all original variables
reintroduces multicollinearity.
6. In Python, which scikit-learn workflow correctly performs PCA and regression within a
single pipeline to avoid data leakage?
[ ] A. Fit PCA on full data, then split and fit regression.
[ ] B. Split data, fit PCA on training set, transform both sets, then fit regression.
[ ] C. Fit regression first, then apply PCA to residuals.
[ ] D. Standardize entire dataset, then split and fit PCA and regression.
CORRECT: B. Split data, fit PCA on training set, transform both sets, then fit regression.
To prevent leakage, PCA must be fit only on training data and then applied to test data. Fitting on full data or
standardizing before split leaks information; fitting regression before PCA is illogical.
HIGHP - Page 3 of 18
Analysis (PCA) | Complete Regression
Report, Python Modeling & Results |
2026/2027 {JUST RELEASED} Updated
Practice Question Bank & Answer Rationale
THIS PACK CONTAINS
50 exam-style multiple-choice questions. Each question includes the correct answer and a complete rationale
for focused revision.
COVERAGE
This exam assesses graduate-level mastery of principal component analysis (PCA) theory, dimensionality
reduction, regression modeling, and Python implementation for a complete analytics report. It covers variance
explained, component selection, standardization, multicollinearity resolution, model diagnostics, and results
communication aligned with WGU D600 Exam 3 (2026/2027).
STUDY GUIDE
Recommended duration: 3 hours. Passing target: 80%.
HIGHP - Page 1 of 18
,1. In a PCA-based regression for home prices, which variable set is most appropriate for
PCA?
[ ] A. SquareFootage, NumBedrooms, BackyardSpace, CrimeRate, SchoolRating, AgeOfHome,
DistanceToCityCenter, EmploymentRate, PropertyTaxRate, RenovationQuality, TransportAccess,
PreviousSalePrice
[ ] B. SquareFootage, NumBedrooms, BackyardSpace, CrimeRate, SchoolRating, AgeOfHome,
DistanceToCityCenter, EmploymentRate, PropertyTaxRate, RenovationQuality, TransportAccess
[ ] C. SquareFootage, NumBedrooms, BackyardSpace, CrimeRate, SchoolRating, AgeOfHome,
DistanceToCityCenter, EmploymentRate, PropertyTaxRate, RenovationQuality, TransportAccess,
PreviousSalePrice, ZipCode
[ ] D. SquareFootage, NumBedrooms, BackyardSpace, CrimeRate, SchoolRating, AgeOfHome,
DistanceToCityCenter, EmploymentRate, PropertyTaxRate, RenovationQuality, TransportAccess,
PreviousSalePrice, SaleDate
CORRECT: B. SquareFootage, NumBedrooms, BackyardSpace, CrimeRate, SchoolRating,
AgeOfHome, DistanceToCityCenter, EmploymentRate, PropertyTaxRate, RenovationQuality,
TransportAccess
PCA should be applied to continuous explanatory variables only; PreviousSalePrice is likely the target or a
leaky variable, and ZipCode/SaleDate are non-continuous. Option B correctly excludes PreviousSalePrice
and non-continuous variables, while A includes the target, and C/D add categorical or date variables
inappropriate for PCA.
2. After standardizing 12 housing variables and running PCA, the first four components
explain 85% of variance. What is the most defensible next step for regression?
[ ] A. Retain all 12 original variables to preserve interpretability.
[ ] B. Use the first four principal components as predictors in a linear regression.
[ ] C. Discard PCA and use ridge regression instead.
[ ] D. Retain components until 99% variance is explained.
CORRECT: B. Use the first four principal components as predictors in a linear regression.
Retaining components that cumulatively explain a substantial portion (e.g., 85%) is a common and defensible
rule for dimensionality reduction. Using all original variables defeats the purpose of PCA, switching to ridge
ignores the PCA goal, and requiring 99% may retain too many components and overfit.
3. Which statement best justifies standardizing variables before PCA in the housing price
analysis?
[ ] A. Standardization ensures all variables have equal weight regardless of scale.
[ ] B. Standardization increases the variance explained by the first component.
[ ] C. Standardization removes multicollinearity among variables.
[ ] D. Standardization makes the data normally distributed.
HIGHP - Page 2 of 18
, CORRECT: A. Standardization ensures all variables have equal weight regardless of scale.
PCA is sensitive to scale; standardizing to mean 0 and SD 1 gives each variable equal influence. It does not
inherently increase variance explained, remove multicollinearity (PCA does that), or normalize distributions.
4. In the regression report, the PCA-transformed predictors are orthogonal. What does this
imply about the regression coefficients?
[ ] A. Coefficients are unaffected by multicollinearity and are stable.
[ ] B. Coefficients are biased due to omitted variables.
[ ] C. Coefficients cannot be interpreted as original variable effects.
[ ] D. Coefficients are all statistically insignificant.
CORRECT: A. Coefficients are unaffected by multicollinearity and are stable.
Orthogonal components eliminate multicollinearity, yielding stable coefficient estimates. However, they are
not directly interpretable as original variable effects (so C is partially true but not the best answer), and they
are not biased or necessarily insignificant.
5. A PCA regression yields R² = 0.78 on training data but 0.45 on test data. Which action
best addresses this?
[ ] A. Add more principal components to the model.
[ ] B. Reduce the number of principal components or apply regularization.
[ ] C. Remove the standardization step.
[ ] D. Use all original variables instead of components.
CORRECT: B. Reduce the number of principal components or apply regularization.
The large gap indicates overfitting; reducing components or regularizing improves generalization. Adding
components worsens overfitting, removing standardization is inappropriate, and using all original variables
reintroduces multicollinearity.
6. In Python, which scikit-learn workflow correctly performs PCA and regression within a
single pipeline to avoid data leakage?
[ ] A. Fit PCA on full data, then split and fit regression.
[ ] B. Split data, fit PCA on training set, transform both sets, then fit regression.
[ ] C. Fit regression first, then apply PCA to residuals.
[ ] D. Standardize entire dataset, then split and fit PCA and regression.
CORRECT: B. Split data, fit PCA on training set, transform both sets, then fit regression.
To prevent leakage, PCA must be fit only on training data and then applied to test data. Fitting on full data or
standardizing before split leaks information; fitting regression before PCA is illogical.
HIGHP - Page 3 of 18
