CENG — Student Risk Analysis

Overview

Model Accuracy

68.8%

Binary logistic regression · 68 correct of 66 observed

Nagelkerke R²

.270 / .200

Nagelkerke / Cox & Snell · Explains 27.0% of variance

Analyzed Cohort

66 / 79

13 excluded — 1–2 semester registrations only

Students with only 1–2 semesters of data were excluded as insufficient longitudinal records preclude reliable risk classification.

Top Risk Factor

CSCI250

Spearman ρ = 0.430, p < .001 · Highest positive correlation

Interactive Tool

Interactive Risk Predictor

Simulate individual student risk odds using the binary regression model

ExpB: EENG300 = 2.716 · PPT = .968

Physics Placement Test (PPT) Score 40

0Threshold (40)100

Repeating EENG300? Primary bottleneck · ExpB 2.716, p = .024

Each additional PPT point above 40 reduces risk odds by 3.2%. Repeating EENG300 multiplies odds by 2.7×.

Risk Multiplier

1.00×

Baseline

Statistical Analysis

Placement Test Discrepancies

Independent t-test · mean scores by risk group

Risk Factor Correlation Map

Spearman ρ · green = protective, red = risk

Course Repetition Analysis

Average registrations per student · all shown courses statistically significant (p < .05)

Regression Model

Omnibus Tests of Model Coefficients

	χ²	df	Sig.
Step / Block / Model	14.282	2	<.001 ***

The overall model is statistically significant — EENG300 and PPT jointly predict risk status beyond chance level.

Model Summary

−2 Log Likelihood72.177

Cox & Snell R².200

Nagelkerke R².270

Variance explained27.0%

Estimation terminated at iteration 5; parameter changes < .001.

Cohort Composition

Total enrolled79

Included in analysis66

Excluded13

Demographics (χ²)Non-significant

Excluded students registered for only 1–2 semesters — insufficient longitudinal data for classification.

Variables in the Equation

Variable	B	S.E.	Wald	df	Sig.	Exp(B)
EENG300 (repeat)	.999	.441	5.124	1	.024 *	2.716
PPT score	−.032	.014	5.550	1	.018 *	.968
Constant	.485	.867	.313	1	.576	1.625

* p < .05 — Significant predictor n.s. — Not significant

Highlighted rows: predictors retained by the model. B = log-odds coefficient; Exp(B) = odds ratio. All remaining Spearman-correlated courses were not significant in the regression.

Classification Table — Cut value 0.500

Predicted
Not At Risk

Predicted
At Risk

Observed
Not At Risk

18True Negative

8False Positive

Observed
At Risk

12False Negative

26True Positive

Not At Risk accuracy69.2%

At Risk accuracy68.4%

Overall accuracy 68.8%

Insights & Interventions

The EENG300 Bottleneck

Repeating EENG300 increases "At Risk" odds by 2.7× (ExpB: 2.716, p = .024) — the single largest controllable risk factor in the model.

PPT as a Protective Factor

Each additional PPT point above 40 reduces risk odds by 3.2% (ExpB: .968, p = .018). A score of 60 implies ~46% lower odds vs. threshold.

Recommended Interventions

Mandate academic advising for any student registering for CSCI250 or EENG300 a second time.

Flag freshmen scoring below 40 on the PPT for a pre-semester physics bridge programme.

Deprioritise high school background in intake risk assessment — statistically non-significant (p = .229–.418).