WGU D600 – Statistical Data Mining Task 3 | Principal Component Analysis (PCA) & Regression | Complete Performance Assessment with Python, Scree Plot, Loadings & Model Evaluation | 2026 Updated

WGU D600 | Task 3 | September 2026




WGU D600 - Statistical Data Mining
Task 3: Principal Component Analysis &
Regression
Performance Assessment Model | September 2026 Alignment



Scenario
A residential property analytics team needs a parsimonious model for predicting sale price from correlated housing
and neighborhood measurements. Principal component analysis is used to transform seven standardized continuous
predictors into orthogonal components before linear regression.



Scope and integrity note
This document is a professionally completed model based on public information about the current D600 course and
Task 3 workflow. WGU learner-portal instructions, assigned data, repository history, and evaluator decisions are not
public and can vary. The representative dataset and outputs here are original and are not presented as WGU-issued
data or as an evaluator-certified pass.




Prepared as a complete worked PCA-to-regression analysis




1

, WGU D600 | Task 3 | September 2026

Executive Summary
This assessment model demonstrates dimensionality reduction with principal component analysis (PCA) followed
by multiple linear regression. The business objective is to predict residential sale price while reducing
redundancy among correlated housing and neighborhood variables. Seven continuous predictors are
standardized, transformed into orthogonal principal components, and evaluated using the Kaiser criterion and a
scree plot.

The Kaiser criterion retains 3 components: PC1, PC2, PC3. Together they explain 87.76% of the predictor variance.
An 80/20 train-test split is created from the retained components and the unstandardized target variable, Price.
The regression model uses the retained PCs only. All retained components are statistically significant at alpha =
0.05, so the backward-elimination check removes no components.

The final training model explains 89.9% of price variation (adjusted R-squared = 0.899) with training RMSE of
$36,389. On the held-out test set, R-squared is 0.903 and RMSE is $37,170. Residual diagnostics support
independence, approximate normality, and constant variance. The model is therefore useful for portfolio-level
estimation and screening, while individual property valuation should still incorporate local comparables and
features not represented in the dataset.

Submission Components Represented
 Research question and analysis goal for linear regression after PCA.
 Rationale for PCA, expected analytical outcomes, and PCA assumption discussion.
 Continuous-variable identification, standardization, and descriptive statistics.
 Complete loading matrix, eigenvalues, scree plot, retention decision, and retained-component variance.
 Train/test datasets containing only retained components plus the original target.
 OLS regression, optimization check, coefficient interpretation, equation, MSE and test-set performance.
 Regression assumption checks, limitations, recommendations, reproducibility steps, and Python code excerpt.
 Checklist for repository URL, commit evidence, data files, and screen-recorded code demonstration that must
be created in the learner's own environment.




2

, WGU D600 | Task 3 | September 2026

1. Current Course Alignment
Western Governors University lists D600 - Statistical Data Mining as a 3-CU course in the September 2026
institutional catalog. The catalog describes the course as covering data preparation, regression, dimensional
reduction, supervised prediction/classification, and unsupervised methods. D600 remains part of the M.S. Data
Analytics curriculum as of September 2026 (Western Governors University, 2026).

Public 2025 course portfolios and learner reports consistently identify Task 3 as a PCA-to-linear-regression
assessment. The most current public evidence also indicates that retained principal components should be
carried into the regression stage rather than performing PCA only as a side analysis. Because exact learner-portal
wording is not publicly available, this report follows the most conservative interpretation: PCA drives the
predictor set used in regression.

Important implementation choice
PCA is fitted to standardized predictor variables only. Price is not standardized and is not included in the PCA. After
the retained components are identified, the transformed component scores are paired with the original Price target
for regression. This preserves price interpretation in dollars.




2. Scenario, Research Question, and Goal
Scenario. HomeValue Analytics supports a regional property company that wants a repeatable way to estimate
residential sale prices. The original predictor set contains correlated measures of property size, neighborhood
quality, location, age, and renovation condition. Correlation can make coefficient-level interpretation unstable
and can add unnecessary dimensionality. PCA is used to consolidate shared variance before prediction.

Research question. To what extent can a reduced set of principal components derived from standardized
continuous housing predictors explain and predict residential sale price?

Analysis goal. The goal is to transform the continuous predictors into orthogonal principal components, retain
only components that explain meaningful variance, use the retained components in linear regression, evaluate
predictive error on unseen observations, and provide a defensible business recommendation based on the
resulting model.




3