WGU D600 - Statistical Data Mining | Task 1
WGU D600
Statistical Data Mining
TASK 1: MULTIPLE LINEAR REGRESSION ANALYSIS
Completed Performance Assessment Model | September 2026
Housing Sale Price Prediction Using Multiple Linear Regression
Scope and alignment
This document is a completed model submission built from WGU's current D600 course description and
publicly documented 2025-2026 Task 1 workflow. WGU does not publish the complete live PA rubric or
proprietary dataset openly. The analysis therefore uses a representative housing dataset and mirrors the
documented requirements: a business/research question, at least three predictors, exploratory analysis,
train/test modeling, multiple linear regression, model reduction, prediction error, regression equation,
assumption checks, interpretation, recommendations, and reproducible Python code. The exact rubric shown
in the learner portal remains controlling.
Prepared as an original academic model. No claim is made that this is an official WGU exemplar or an evaluator-issued pass.
September 2026 aligned model submission | Page 1
, WGU D600 - Statistical Data Mining | Task 1
Executive Summary
The purpose of this analysis is to determine whether a multiple linear regression model can provide a
defensible estimate of residential sale price from structural and neighborhood characteristics. The
representative housing scenario is consistent with public D600 learner reports that identify housing as a
current data option and describe Task 1 as a multiple linear regression assessment. The analysis uses
1,200 residential property records and an 80/20 train-test split.
The initial model used seven independent variables. One predictor, NumBedrooms, was not statistically
significant after the effects of living area and bathroom count were controlled (p = 0.518); it was removed
using backward elimination. The final six-predictor model was statistically significant overall (F = 611.7, p
< .001), explained 79.4% of training-set variation in price, and achieved a test-set R-squared of 0.822. Test
RMSE was $55,466 and test MAE was $45,006.
The diagnostics support use of the model for instructional decision support: residual independence is
acceptable (Durbin-Watson near 2), the Breusch-Pagan test does not indicate statistically significant
heteroscedasticity, residual normality is reasonable, and centered predictor VIF values are below 2. The
model is therefore suitable as a baseline pricing model for the scenario, while the error magnitude and
synthetic/representative nature of the dataset mean it should not be used as an automated appraisal
engine without validation on real market data.
Metric Result Interpretation
Training R-squared 0.794 Substantial explanatory power for a
baseline linear model
Adjusted R-squared 0.793 Little evidence of unnecessary
complexity
Test R-squared 0.822 Model generalizes comparably on
held-out data
Test RMSE $55,466 Typical squared-error-sensitive
prediction error
Test MAE $45,006 Average absolute error in practical
dollar terms
Breusch-Pagan p-value 0.148 No significant heteroscedasticity
detected at alpha=.05
Durbin-Watson 2.014 Residual independence is consistent
with expectations
Jarque-Bera p-value 0.349 Residual normality is not rejected at
alpha=.05
Current Course and Task Verification
WGU's 2026 institutional catalog lists D600 Statistical Data Mining as a 3-CU course following D599 Data
Preparation and Exploration. The official description explicitly includes statistics, data preparation,
regression, dimensional reduction, supervised classification/prediction, and unsupervised modeling.
Public D600 learner reports consistently identify three performance assessments: linear regression,
logistic regression, and principal component analysis.
September 2026 aligned model submission | Page 2
, WGU D600 - Statistical Data Mining | Task 1
2026 version-control change
A January 2026 learner report documents a revised D600 version in which the earlier requirement to create
a sequence of GitLab commits for each rubric subpart was removed. Learners in that version still reported
placing the finished Python code in GitLab. Because WGU can assign different assessment versions, the live
learner-portal rubric should be checked before submission.
1. Business Scenario
A regional residential-property analytics team supports acquisition, pricing, and market-review decisions.
Leadership wants an interpretable statistical model that estimates sale price from characteristics known
before a transaction closes. An interpretable linear model is appropriate because stakeholders need both a
numeric prediction and a transparent explanation of how each predictor changes the expected price while
other variables are held constant.
The representative data contain property size, bedrooms, bathrooms, age, neighborhood crime, school
quality, distance from the city center, and observed sale price. The dependent variable is continuous, the
predictors are quantitative, and the business question concerns prediction and explanation. These
characteristics make multiple linear regression an appropriate first model.
2. Research Question and Analysis Goal
2.1 Research Question
Research question: To what extent can living area, number of bedrooms, number of bathrooms, property
age, neighborhood crime, school rating, and distance from the city center predict residential sale price?
2.2 Analysis Goal
The goal is to construct, evaluate, and reduce a multiple linear regression model that provides an
interpretable estimate of sale price and quantifies out-of-sample prediction error. A secondary goal is to
identify which predictors remain useful after the other variables are controlled.
2.3 Expected Organizational Benefit
Support preliminary pricing and acquisition screening with an evidence-based baseline estimate.
Identify variables with meaningful positive or negative associations with sale price.
Quantify expected prediction error so stakeholders understand uncertainty.
Provide a transparent benchmark against which more complex future models can be compared.
3. Data, Variables, and Preparation
3.1 Analytic Dataset
The analysis uses a representative housing dataset of 1,200 property-level observations. The file is treated
as an already assembled analytic table. Data-quality checks confirm no missing values and no duplicate
September 2026 aligned model submission | Page 3
WGU D600
Statistical Data Mining
TASK 1: MULTIPLE LINEAR REGRESSION ANALYSIS
Completed Performance Assessment Model | September 2026
Housing Sale Price Prediction Using Multiple Linear Regression
Scope and alignment
This document is a completed model submission built from WGU's current D600 course description and
publicly documented 2025-2026 Task 1 workflow. WGU does not publish the complete live PA rubric or
proprietary dataset openly. The analysis therefore uses a representative housing dataset and mirrors the
documented requirements: a business/research question, at least three predictors, exploratory analysis,
train/test modeling, multiple linear regression, model reduction, prediction error, regression equation,
assumption checks, interpretation, recommendations, and reproducible Python code. The exact rubric shown
in the learner portal remains controlling.
Prepared as an original academic model. No claim is made that this is an official WGU exemplar or an evaluator-issued pass.
September 2026 aligned model submission | Page 1
, WGU D600 - Statistical Data Mining | Task 1
Executive Summary
The purpose of this analysis is to determine whether a multiple linear regression model can provide a
defensible estimate of residential sale price from structural and neighborhood characteristics. The
representative housing scenario is consistent with public D600 learner reports that identify housing as a
current data option and describe Task 1 as a multiple linear regression assessment. The analysis uses
1,200 residential property records and an 80/20 train-test split.
The initial model used seven independent variables. One predictor, NumBedrooms, was not statistically
significant after the effects of living area and bathroom count were controlled (p = 0.518); it was removed
using backward elimination. The final six-predictor model was statistically significant overall (F = 611.7, p
< .001), explained 79.4% of training-set variation in price, and achieved a test-set R-squared of 0.822. Test
RMSE was $55,466 and test MAE was $45,006.
The diagnostics support use of the model for instructional decision support: residual independence is
acceptable (Durbin-Watson near 2), the Breusch-Pagan test does not indicate statistically significant
heteroscedasticity, residual normality is reasonable, and centered predictor VIF values are below 2. The
model is therefore suitable as a baseline pricing model for the scenario, while the error magnitude and
synthetic/representative nature of the dataset mean it should not be used as an automated appraisal
engine without validation on real market data.
Metric Result Interpretation
Training R-squared 0.794 Substantial explanatory power for a
baseline linear model
Adjusted R-squared 0.793 Little evidence of unnecessary
complexity
Test R-squared 0.822 Model generalizes comparably on
held-out data
Test RMSE $55,466 Typical squared-error-sensitive
prediction error
Test MAE $45,006 Average absolute error in practical
dollar terms
Breusch-Pagan p-value 0.148 No significant heteroscedasticity
detected at alpha=.05
Durbin-Watson 2.014 Residual independence is consistent
with expectations
Jarque-Bera p-value 0.349 Residual normality is not rejected at
alpha=.05
Current Course and Task Verification
WGU's 2026 institutional catalog lists D600 Statistical Data Mining as a 3-CU course following D599 Data
Preparation and Exploration. The official description explicitly includes statistics, data preparation,
regression, dimensional reduction, supervised classification/prediction, and unsupervised modeling.
Public D600 learner reports consistently identify three performance assessments: linear regression,
logistic regression, and principal component analysis.
September 2026 aligned model submission | Page 2
, WGU D600 - Statistical Data Mining | Task 1
2026 version-control change
A January 2026 learner report documents a revised D600 version in which the earlier requirement to create
a sequence of GitLab commits for each rubric subpart was removed. Learners in that version still reported
placing the finished Python code in GitLab. Because WGU can assign different assessment versions, the live
learner-portal rubric should be checked before submission.
1. Business Scenario
A regional residential-property analytics team supports acquisition, pricing, and market-review decisions.
Leadership wants an interpretable statistical model that estimates sale price from characteristics known
before a transaction closes. An interpretable linear model is appropriate because stakeholders need both a
numeric prediction and a transparent explanation of how each predictor changes the expected price while
other variables are held constant.
The representative data contain property size, bedrooms, bathrooms, age, neighborhood crime, school
quality, distance from the city center, and observed sale price. The dependent variable is continuous, the
predictors are quantitative, and the business question concerns prediction and explanation. These
characteristics make multiple linear regression an appropriate first model.
2. Research Question and Analysis Goal
2.1 Research Question
Research question: To what extent can living area, number of bedrooms, number of bathrooms, property
age, neighborhood crime, school rating, and distance from the city center predict residential sale price?
2.2 Analysis Goal
The goal is to construct, evaluate, and reduce a multiple linear regression model that provides an
interpretable estimate of sale price and quantifies out-of-sample prediction error. A secondary goal is to
identify which predictors remain useful after the other variables are controlled.
2.3 Expected Organizational Benefit
Support preliminary pricing and acquisition screening with an evidence-based baseline estimate.
Identify variables with meaningful positive or negative associations with sale price.
Quantify expected prediction error so stakeholders understand uncertainty.
Provide a transparent benchmark against which more complex future models can be compared.
3. Data, Variables, and Preparation
3.1 Analytic Dataset
The analysis uses a representative housing dataset of 1,200 property-level observations. The file is treated
as an already assembled analytic table. Data-quality checks confirm no missing values and no duplicate
September 2026 aligned model submission | Page 3
