index Introduction
Concepts, Techniques and Tools

Relationship & Driver Analysis

What factors drive results?
Set 04 · Level 4 of the Analytics Ladder

What Factors Drive the Results You See?

Moving from describing what happened and when — to understanding which variables are responsible.

4Analytics Ladder · Level 04 of 10

Sets 01–03 described data, diagnosed failures, and projected trends. Set 04 answers the next logical question: which variables actually drive the outcomes we care about? If revenue is growing, what is driving it — price, volume, mix, or something else? If customer satisfaction is declining, which factors explain the most variation?

The techniques in this set quantify relationships — attaching numbers to the question "how much does Y change when X changes?" This shifts analysis from observation to explanation.

Correlation shows that two things move together. Regression quantifies how much one changes when the other does. Driver analysis ranks which inputs matter most.
8
Techniques
Key Statistic
variance explained
β
Regression
coefficient
4
Tools Covered
Technique 01

Pearson Correlation

Measures the strength and direction of the linear relationship between two continuous variables. Produces r between −1 and +1.

Technique 02

Spearman Correlation

Non-parametric alternative to Pearson. Uses ranks instead of raw values. Appropriate when data is ordinal or non-normally distributed.

Technique 03

Simple Linear Regression

Models the relationship between one predictor (X) and one outcome (Y). Produces a slope (β) and intercept (α): Y = α + βX.

Technique 04

Multiple Linear Regression

Extends simple regression to multiple predictors simultaneously. Estimates the unique contribution of each predictor while controlling for the others.

Technique 05

Partial Correlation

Measures the relationship between two variables after removing the effect of one or more control variables. Isolates the "pure" relationship.

Technique 06

Elasticity Analysis

Measures the % change in Y for a 1% change in X. Used extensively in pricing, demand analysis, and economic modelling.

Technique 07

Driver Analysis (Relative Weight)

Ranks independent variables by their proportional contribution to R². Answers: "Which driver explains the most variation in the outcome?"

Technique 08

Path Analysis (IBM AMOS)

Tests direct and indirect relationships between variables. Goes beyond regression by modelling causal chains — X affects Y both directly and through mediating variable M.

⚠ Still not causation

All regression and correlation techniques identify statistical associations — even when they are strong, significant, and robust. Causal interpretation requires the disciplined reasoning covered in Sets 05 (Multivariate Analysis) and the analytical thinking programme's Modules 04–05. Use these techniques to quantify relationships; use Sets 05+ to defend causal claims.

Techniques 01 & 02

Pearson & Spearman Correlation

The first question in any driver analysis: do these two variables move together — and how strongly?

Pearson Correlation Coefficient (r)

Pearson r measures the strength and direction of the linear relationship between two continuous variables. It ranges from −1 (perfect negative) to +1 (perfect positive), with 0 indicating no linear relationship.

r valueStrengthBusiness interpretationTypical example
+0.9 to +1.0Very strong positiveX almost always rises when Y risesAd spend and revenue in a growing market
+0.6 to +0.9Strong positiveClear positive pattern — X is likely a meaningful driverTraining hours and performance scores
+0.3 to +0.6Moderate positiveSome positive relationship; other factors also matterCustomer age and product spend
−0.3 to +0.3Weak or no relationshipLittle evidence of linear associationShoe size and job performance
−0.6 to −0.3Moderate negativeX tends to rise as Y fallsWait time and satisfaction score
−0.9 to −0.6Strong negativeStrong inverse relationshipMachine age and output quality
⚠ When Pearson is not appropriate

Pearson assumes both variables are continuous and normally distributed, and measures only linear relationships. Use Spearman instead when: data is ordinal (satisfaction ratings, rankings) · data is not normally distributed · the relationship is monotonic but non-linear (always increasing or always decreasing, but not at a constant rate).

Spearman Rank Correlation (ρ)

Spearman's ρ (rho) works the same way as Pearson — also ranges from −1 to +1 — but instead of using raw values it converts values to ranks first, then calculates the correlation of the ranks. This makes it robust to outliers and non-normal distributions.

Business Example
When to choose Spearman over Pearson

An HR analyst wants to understand whether employee satisfaction rank correlates with productivity rank. Because satisfaction is measured on a 1–5 Likert scale (ordinal data) and productivity has a few extreme outliers, Spearman is the correct choice. The ranks are stable and the outliers do not distort the correlation. Pearson would give a misleadingly weak result because of the outliers.

What is R² (Coefficient of Determination)?

R² = r × r. If r = 0.80, then R² = 0.64. This means 64% of the variation in Y is explained by the linear relationship with X. The remaining 36% is explained by other factors not in the model. R² is the single most important number to understand in regression output.

r valueR² valuePlain interpretation
0.900.81X explains 81% of the variation in Y. Very high — most of Y's movement is accounted for by X.
0.700.49X explains 49% of variation. Substantial — but half of Y's variation comes from other factors.
0.500.25X explains only 25%. Real relationship, but most of Y is explained by things not in this model.
0.300.09X explains only 9%. Weak — X matters but barely. Probably not the key driver.
Technique 03

Simple Linear Regression

One predictor. One outcome. How much does Y change for each unit increase in X?

Correlation tells you that two variables are related. Regression quantifies the relationship. Simple linear regression fits the best possible straight line through a scatter of points — the line that minimises the total prediction error across all observations.

Y = α + β × X   +   ε
Y
Outcome variable
(what we want to predict)
α
Intercept
(value of Y when X=0)
β
Slope coefficient
(change in Y per unit of X)
ε
Error term
(unexplained variation)
Business Example
Sales Staff Training Hours vs Revenue per Rep

Regression of monthly revenue per sales rep (Y) on training hours completed (X) produces: Revenue = $42,000 + $1,850 × Training Hours. R² = 0.67.

Interpretation: Each additional training hour is associated with $1,850 more monthly revenue per rep. Training hours explain 67% of the variation in revenue across reps. A rep with zero training is predicted to generate $42,000 (the base intercept). A rep with 20 training hours: $42,000 + (20 × $1,850) = $79,000.

Caveat: this is an association, not a proven causal effect. High performers may both train more AND generate more revenue for independent reasons.

Reading regression output — the key numbers

Output elementWhat it meansRule of thumb
R² (R-squared)% of outcome variance explained by the modelHigher is better. Context-dependent: R²=0.30 may be excellent for behaviour prediction; R²=0.30 is poor for cost modelling.
Adjusted R²R² penalised for adding extra predictors that don't helpAlways use Adjusted R² when comparing models with different numbers of predictors.
β coefficientChange in Y per 1-unit increase in XMust be interpreted in the units of both X and Y. "$1,850 per training hour" is clear. "0.023" is not — always add units.
p-value of βProbability that this coefficient is due to chance if the true relationship is zerop < 0.05 is the conventional threshold (5% chance of false positive). p < 0.01 is stronger. p > 0.05 = not statistically significant.
Standard Error (SE)Uncertainty in the β estimate95% confidence interval ≈ β ± 2×SE. A wide SE means the coefficient is imprecisely estimated.
F-statistic (overall model)Tests whether the model explains significantly more than chancep-value of F should be < 0.05 for the model to be considered statistically useful.
Residual plotThe unexplained variation — should look randomIf residuals show a pattern (curved, fan-shaped), the model is misspecified — try a transformation or a different functional form.
Technique 04

Multiple Linear Regression

Multiple predictors, one outcome — estimating each driver's unique contribution while controlling for the others.

Multiple regression extends simple regression to include several predictors simultaneously. This is more realistic — most business outcomes are driven by more than one factor. The critical advantage is control: each coefficient represents the change in Y per unit of that predictor, holding all other predictors constant.

Y = α + β₁X₁ + β₂X₂ + β₃X₃ + ε
Business Example
What drives customer lifetime value?

A subscription company runs multiple regression with CLV ($) as Y and three predictors: Onboarding Score (X₁), Number of Features Used (X₂), Support Tickets Raised (X₃).

Output: CLV = $800 + $42×Onboarding + $95×Features − $180×Support Tickets. Adjusted R² = 0.71.

Interpretation: Holding other factors constant, each additional feature used is associated with $95 more CLV. Each support ticket is associated with $180 less CLV. Features used has the largest positive driver effect; support tickets have the largest negative effect. This tells the product team where to invest: feature adoption is worth more than fixing onboarding.

The multicollinearity problem

Multicollinearity occurs when two or more predictors are highly correlated with each other. This makes it difficult for the model to isolate the unique contribution of each — coefficients become unstable and standard errors inflate. It is the most common problem in multiple regression.

VIF valueDiagnosisAction
1.0 – 2.0No multicollinearityNo action needed
2.0 – 5.0Moderate — acceptable in most casesMonitor; coefficients are still interpretable
5.0 – 10.0High — problematicConsider removing one of the correlated predictors or combining them
> 10.0Severe — coefficients are unreliableRemove predictors until VIF is below 10. Do not interpret individual coefficients.
VIF — Variance Inflation Factor

VIF measures how much the variance of a coefficient is inflated due to correlation with other predictors. In Excel: calculate manually as 1/(1−R²j) where R²j is the R² from regressing predictor j on all other predictors. In SPSS: Analyze → Regression → Linear → Statistics → tick Collinearity Diagnostics. In Python: from statsmodels.stats.outliers_influence import variance_inflation_factor.

Standardised coefficients (Beta weights)

Raw regression coefficients (β) cannot be compared directly if predictors are in different units — "training hours" and "onboarding score" are on completely different scales. Standardised coefficients (Beta weights) convert all predictors to the same scale (standard deviations), allowing direct comparison: which predictor has the largest standardised effect on the outcome?

Techniques 05 & 06

Partial Correlation & Elasticity Analysis

Two specialist techniques for more precise driver quantification.

Partial Correlation

Partial correlation measures the relationship between two variables after statistically removing the influence of one or more control variables. It answers: "If I could hold Z constant, what is the true relationship between X and Y?"

Business Example
Advertising spend and revenue — controlling for seasonality

A company finds a strong Pearson r = 0.75 between advertising spend and revenue. But both tend to peak in Q4 (Christmas). Is the correlation real, or is it driven by the shared seasonality?

The partial correlation between advertising and revenue, controlling for the month of year, comes back at r = 0.52. Still meaningful, but lower — confirming that some of the simple correlation was seasonal artefact. The "pure" advertising effect is moderate, not strong.

Elasticity Analysis

Elasticity measures the % change in Y for a 1% change in X. Unlike regression coefficients (which are in raw units), elasticity is unit-free — making it easy to compare the sensitivity of different variables.

Elasticity = (%ΔY) / (%ΔX) = (ΔY/Y) / (ΔX/X)
Elasticity valueInterpretationBusiness meaning
> 1.0 (elastic)Y changes more than proportionally to XA 10% price increase leads to >10% demand drop. Customers are price-sensitive. Revenue falls if you raise price.
= 1.0 (unit elastic)Y changes proportionally to XA 10% price rise leads to exactly 10% demand drop. Revenue is unchanged.
0 to 1.0 (inelastic)Y changes less than proportionally to XA 10% price rise leads to <10% demand drop. Customers are price-insensitive. Revenue increases if you raise price.
< 0 (negative)Y moves in the opposite direction to XStandard demand: as price rises, demand falls. Negative price elasticity is normal.
Business Example
Price elasticity of demand — luxury vs essential goods

A FMCG company tests price changes across two product categories. Bottled water: price elasticity = −0.3 (inelastic). A 10% price increase reduces volume by only 3% — customers still buy it because it's perceived as essential. Revenue rises. Premium snack range: price elasticity = −1.8 (elastic). A 10% price increase reduces volume by 18% — customers switch to alternatives. Revenue falls. Same 10% price increase, very different strategic implications.

How to calculate elasticity from regression

If you run a log-log regression: ln(Y) = α + β·ln(X), the coefficient β is directly the elasticity. A 1% increase in X is associated with a β% change in Y. This is the standard approach in demand modelling and economic analysis.

Techniques 07 & 08

Driver Analysis & Path Analysis

Which predictor matters most — and do variables work directly or through others?

Driver Analysis (Relative Weight Analysis)

Multiple regression tells you whether each predictor is significant. Driver analysis answers a different question: which predictor contributes the most to explaining the outcome — accounting for correlations between predictors?

Standard beta weights from regression can be misleading when predictors are correlated. Relative Weight Analysis (RWA) decomposes R² into the portion attributable to each predictor, even when predictors share variance. The result: a percentage contribution that sums to 100%.

Business Example
Key drivers of Net Promoter Score

A telecom company runs driver analysis on NPS (Y) with four predictors: Product Quality, Price Fairness, Customer Service, and Network Reliability. Relative weight results:

DriverRelative Weight (%)Beta (standardised)Priority
Network Reliability38%0.52🔴 Top priority
Customer Service29%0.41🟡 High priority
Price Fairness20%0.34🟡 Monitor
Product Quality13%0.28🟢 Maintain

Invest in network reliability first — it accounts for 38% of NPS variance. Price is real but less central than reliability or service.

Path Analysis (IBM AMOS)

Path analysis tests whether a variable affects the outcome both directly and through other mediating variables. Standard regression only captures direct effects. Path analysis maps the full causal chain.

Business Example
Does training improve customer satisfaction directly — or only through better service quality?

A simple regression: Training → Customer Satisfaction (β = 0.45, significant). But there may be a mediating path: Training → Service Quality → Customer Satisfaction. Path analysis (AMOS) tests both paths simultaneously:

Results: Training → Customer Satisfaction directly: β = 0.12 (weak, now non-significant). Training → Service Quality → Customer Satisfaction: indirect effect = 0.38 (significant).

Conclusion: Training improves satisfaction primarily through service quality, not directly. Investment in training must be paired with service quality measurement — otherwise it won't move the satisfaction needle.

IBM AMOS — key fit indices to check

After running a path model in AMOS, assess model fit with these indices: CFI > 0.95 (Comparative Fit index — how well model fits vs a null model) · RMSEA < 0.06 (Root Mean Square Error of Approximation — lower is better) · SRMR < 0.08 (Standardised Root Mean Residual) · χ² p-value > 0.05 (non-significant means model fits the data). All four should be checked — no single index is sufficient.

Tool Guide 🟢 Start Here

Microsoft Excel

Correlation, simple regression, and multiple regression — all via the Data Analysis ToolPak.

🟢 Start HereCorrelation Matrix
01
Organise your data
Place all numeric variables in adjacent columns with header row. No blank columns between variables.
02
Run correlation
Data → Data Analysis → Correlation → OK → Input Range: select all numeric columns including headers → tick Labels in First Row → choose output location → OK.
03
Read the matrix
Each cell shows the Pearson r between two variables. The diagonal is always 1.0 (a variable perfectly correlates with itself). Focus on the upper or lower triangle — they are identical. Flag cells where |r| > 0.5 for further investigation.
Quick interpretation rule
|r| > 0.7 = strong · 0.4–0.7 = moderate · <0.4 = weak. Highlight strong correlations with conditional formatting for quick visual scan.
🟢 Start HereRegression (Simple & Multiple)
01
Prepare your data
Y (outcome) in one column. X predictors in adjacent columns. All data must be numeric — no text values. Ensure no blank rows within the data range.
02
Launch regression
Data → Data Analysis → Regression → OK
03
Configure
Input Y Range: select Y column including header · Input X Range: select all X columns including headers (multiple columns = multiple regression) · tick Labels · tick Confidence Level (95%) · tick Residuals and Residual Plots · choose output location → OK.
04
Read the output — three tables
Table 1 — Regression Statistics: R Square (variance explained) · Adjusted R Square · Standard Error (model precision).
Table 2 — ANOVA: F-statistic and its p-value. p<0.05 = model is statistically significant overall.
Table 3 — Coefficients: Intercept and each X coefficient (β) · Standard Error · t Stat · P-value · 95% Confidence Interval. Look for p-values <0.05.
What to report
"Revenue = $42,000 + $1,850 × Training Hours (β = $1,850, p = 0.002, 95% CI: $1,210–$2,490). R² = 0.67 — training hours explain 67% of revenue variance across reps."
05
Check the residual plot
Excel produces a Residual vs X plot. It should look like a random scatter (no pattern). A U-shape suggests a non-linear relationship — add a squared term. A funnel shape suggests heteroscedasticity — consider log-transforming Y.
🟢 Start HereUseful Regression Formulas
FormulaPurpose
=CORREL(A2:A100, B2:B100)Pearson r between two columns
=RSQ(known_y, known_x)R² for simple regression
=SLOPE(known_y, known_x)β coefficient (slope)
=INTERCEPT(known_y, known_x)α (intercept)
=FORECAST.LINEAR(x_new, known_y, known_x)Predicted Y for a new X value
=STEYX(known_y, known_x)Standard error of the regression
Tool Guide 🟡 Go Deeper

SPSS & IBM AMOS

Professional regression with full diagnostics in SPSS. Path analysis and structural models in AMOS.

🟡 Go DeeperSPSS — Linear Regression
01
Open the Regression dialog
Analyze → Regression → Linear
02
Add variables
Move your outcome variable to Dependent. Move one or more predictor variables to Independent(s). For multiple regression, add all predictors to the same Independent block.
03
Request statistics
Click Statistics → tick: Estimates · Confidence intervals (95%) · R squared change · Descriptives · Collinearity diagnostics (VIF) · Part and partial correlations → Continue.
04
Request plots
Click Plots → tick *ZPRED on Y-axis, *ZRESID on X-axis (standardised residuals vs predicted) · tick Histogram and Normal probability plot of residuals → Continue → OK.
SPSS output includes
Model Summary (R, R², Adjusted R², Durbin-Watson) · ANOVA table (F, significance) · Coefficients table (B, Std Error, Beta, t, Sig., VIF) · Residual plots for assumption checking
05
Read the Coefficients table
Focus on: B = raw coefficient (in original units) · Beta = standardised coefficient (for comparing drivers) · Sig. = p-value (want <0.05) · Tolerance/VIF = check for multicollinearity (VIF <5 acceptable).
🟡 Go DeeperSPSS — Partial Correlation
01
Open Partial Correlations
Analyze → Correlate → Partial
02
Specify variables
Move the two variables whose relationship you want to examine into the Variables box. Move the control variable(s) into the Controlling for box → OK.
03
Interpret
Compare the partial correlation to the zero-order (Pearson) correlation. A large drop when controlling = the original correlation was partly spurious. A small drop = the relationship is relatively direct.
🟡 Go DeeperIBM AMOS — Path Analysis

IBM AMOS (Analysis of Moment Structures) is a separate application that connects to SPSS. It uses a visual drag-and-drop interface to build path diagrams — no code required.

01
Open your data in SPSS, then launch AMOS
With your data file open in SPSS: Analyze → IBM SPSS Amos (or launch from the Start menu). In AMOS, click File → Data Files → File Name and link to your SPSS dataset.
02
Draw your path model
Use the toolbar to draw rectangles (observed variables — variables in your dataset) and ovals (latent/error variables). Draw single-headed arrows for directional effects (X → Y) and double-headed arrows for correlations. Name each rectangle with the exact variable name from your dataset.
03
Add error terms
Every endogenous variable (one with an arrow pointing TO it) needs an error term. Click the "Add a unique variable to an existing variable" button and click each endogenous variable. AMOS adds an error oval automatically.
04
Run the model
Analyze → Calculate Estimates (or press F9). AMOS estimates all path coefficients simultaneously using Maximum Likelihood estimation.
05
Read model fit and path coefficients
In the Output: Standardised estimates tab shows path coefficients (β) on each arrow · Model Fit tab shows CFI, RMSEA, SRMR, χ² · Estimates tab shows unstandardised coefficients with standard errors and p-values.
Acceptable fit criteria
CFI ≥ 0.95 · RMSEA ≤ 0.06 · SRMR ≤ 0.08 · χ² p > 0.05 (non-significant = model fits data)
Tool Guide 🔵 Full Power

Python — statsmodels, scipy & pingouin

Full regression pipeline — from correlation to multiple regression, VIF, and standardised coefficients.

🔵 Full PowerCorrelation Analysis
# Pearson and Spearman correlation import pandas as pd from scipy import stats import seaborn as sns import matplotlib.pyplot as plt # Full correlation matrix (Pearson) corr_matrix = df.select_dtypes(include='number').corr() sns.heatmap(corr_matrix, annot=True, fmt='.2f', cmap='RdYlGn', center=0) plt.show() # Single Pearson r with p-value r, p = stats.pearsonr(df['training_hours'], df['revenue']) print(f"Pearson r = {r:.3f}, p = {p:.4f}, R² = {r**2:.3f}") # Spearman (for ordinal or non-normal data) rho, p_s = stats.spearmanr(df['satisfaction_rank'], df['productivity_rank']) print(f"Spearman ρ = {rho:.3f}, p = {p_s:.4f}") # Partial correlation (controlling for a third variable) # pip install pingouin import pingouin as pg partial = pg.partial_corr(data=df, x='ad_spend', y='revenue', covar='month') print(partial)
🔵 Full PowerSimple & Multiple Regression
# Simple regression with full output import statsmodels.api as sm X = sm.add_constant(df['training_hours']) # adds intercept y = df['revenue'] model = sm.OLS(y, X).fit() print(model.summary()) # full output: R², F-stat, coefficients, p-values # Multiple regression X_multi = sm.add_constant(df[['onboarding_score', 'features_used', 'support_tickets']]) model_multi = sm.OLS(df['clv'], X_multi).fit() print(model_multi.summary()) # Extract key numbers print(f"R² = {model_multi.rsquared:.3f}") print(f"Adj R² = {model_multi.rsquared_adj:.3f}") print(model_multi.params) # coefficients print(model_multi.pvalues) # p-values print(model_multi.conf_int()) # 95% confidence intervals # Check multicollinearity (VIF) from statsmodels.stats.outliers_influence import variance_inflation_factor vif = pd.DataFrame() vif['Variable'] = X_multi.columns vif['VIF'] = [variance_inflation_factor(X_multi.values, i) for i in range(X_multi.shape[1])] print(vif) # VIF > 10 = severe multicollinearity
🔵 Full PowerStandardised Coefficients & Elasticity
# Standardised coefficients (Beta weights) from sklearn.preprocessing import StandardScaler scaler = StandardScaler() X_std = scaler.fit_transform(df[['onboarding_score','features_used','support_tickets']]) y_std = scaler.fit_transform(df[['clv']]).ravel() X_std_wc = sm.add_constant(X_std) model_std = sm.OLS(y_std, X_std_wc).fit() print("Standardised Beta weights:") print(model_std.params[1:]) # skip constant — these are directly comparable # Price elasticity via log-log regression import numpy as np df['log_price'] = np.log(df['price']) df['log_demand'] = np.log(df['demand']) X_elas = sm.add_constant(df['log_price']) elas_model = sm.OLS(df['log_demand'], X_elas).fit() elasticity = elas_model.params['log_price'] print(f"Price elasticity = {elasticity:.2f}") # Negative = demand falls as price rises (expected) # |elasticity| > 1 = elastic (price-sensitive)
Interactive Practice

Three Simulators

Build intuition for correlation, regression, and multicollinearity through live interaction.

Simulator 1 · Correlation Strength Explorer

Adjust the slider to control the true correlation between two variables. Watch how the scatter plot and r value change — and feel the difference between a "strong" and "weak" relationship.

r = −1.0−0.50+0.5r = +1.0
Pearson r
R² (% variance explained)
Strength
Move the slider to change the true correlation. Click New Sample to regenerate data.

Simulator 2 · Regression Coefficient Reader

Three business scenarios are presented. Read the regression output and answer what the coefficient means in plain language. Tests your ability to translate numbers into business insight.

Scenario 1 of 4
Read the regression output above and select the correct plain-language interpretation.

Simulator 3 · Multicollinearity Impact

Increase the correlation between two predictors and watch what happens to their regression coefficients and VIF. This demonstrates why multicollinearity makes coefficients unreliable.

Increase the correlation between predictors and observe the effect on VIF and coefficient stability.
Best Practice & Assessment

Common Mistakes & Knowledge Check

Six mistakes that turn good data into misleading conclusions — followed by five assessment questions.

Mistake 01

Claiming causation from regression

A significant β coefficient shows a statistical association — not a cause. Omitted variables, reverse causality, and confounding are always possible. Never say "X causes Y" from regression alone.

Mistake 02

Ignoring multicollinearity

When predictors are highly correlated, individual coefficients become unstable and uninterpretable. Always check VIF. If VIF >10, do not interpret the individual coefficients — they are unreliable.

Mistake 03

Comparing raw β coefficients across variables in different units

"Training hours: β=1,850" vs "Satisfaction score: β=420" cannot be compared. $1,850 per hour vs $420 per score point are completely different scales. Always use standardised Beta weights for driver ranking.

Mistake 04

Not checking residuals

A significant model with good R² can still be badly misspecified. If the residuals show a pattern (curve, funnel, clusters), the model is wrong. Always plot residuals before reporting results.

Mistake 05

Overfitting with too many predictors

Adding more predictors always increases R² — even if they are irrelevant. Use Adjusted R² (penalises for extra predictors) and always check out-of-sample performance before acting on a model with many variables.

Mistake 06

Extrapolating beyond the data range

If your training data ranged from 0–20 hours, predicting for 50 hours is unreliable. The regression line is valid only within the observed range. Extrapolation assumes the linear relationship continues indefinitely — which it rarely does.

Set 04 · Key Takeaway
Regression quantifies the relationship. It does not prove what causes what. Always interpret in context — and always check the residuals.

Knowledge Check

Q1 — A regression model gives β = 0.68 with p = 0.001 for the relationship between advertising spend and revenue. R² = 0.46. What does R² = 0.46 mean?
A
Advertising spend explains 46 percentage points of revenue
B
Advertising spend explains 46% of the variation in revenue across observations
C
The model is 46% accurate
D
There is a 46% probability that the relationship is real
R² = 0.46 means advertising spend accounts for 46% of the variance in revenue across the dataset. The remaining 54% is explained by other factors not in the model. It does not mean the model is 46% accurate — that is a different concept (accuracy is for classification models). R² is always between 0 and 1, representing the proportional reduction in unexplained variance.
Q2 — Two predictors in a multiple regression have VIF = 12.4 and VIF = 11.8. What should the analyst do?
A
Report the coefficients as normal — VIF above 10 is common
B
Do not interpret the individual coefficients — the two predictors are severely collinear. Consider removing one or combining them
C
Increase the sample size to reduce VIF
D
Use Spearman correlation instead of regression
VIF above 10 indicates severe multicollinearity — the two predictors are so highly correlated that the model cannot reliably separate their individual effects. The coefficients are unstable (small data changes would produce very different estimates). The solution: remove one of the collinear predictors, combine them into a composite, or use Principal Components (covered in Set 10) to create uncorrelated predictors.
Q3 — A company finds that advertising spend (β = $2,400/unit) and customer service rating (β = $180/unit) both significantly predict CLV. The analyst concludes advertising is a stronger driver. What is missing from this comparison?
A
A significance test between the two coefficients
B
Standardised Beta weights — raw coefficients cannot be compared because advertising spend and service ratings are in different units
C
A larger sample size before comparing
D
A correlation matrix between the two predictors
Advertising spend in dollars vs service rating on a 1–10 scale are completely different units. A β of $2,400 per dollar of spend is meaningless compared to $180 per rating point without knowing the typical range of each variable. Standardised Beta weights (computed by standardising all variables to mean=0, SD=1) put everything on the same scale — then you can legitimately compare which predictor has the larger standardised effect.
Q4 — A product has price elasticity of −0.4. A competitor has price elasticity of −2.1. If both companies raise prices by 15%, what happens to each company's revenue?
A
Both companies lose revenue — higher prices always reduce revenue
B
Company 1 (−0.4, inelastic) gains revenue. Company 2 (−2.1, elastic) loses revenue significantly
C
Both companies gain revenue — prices are higher
D
Company 1 loses revenue. Company 2 gains revenue
Company 1: elasticity −0.4 (inelastic). A 15% price rise → only 6% demand drop (0.4×15%). Revenue = Price × Volume: price up 15%, volume down only 6% → net revenue gain. Company 2: elasticity −2.1 (elastic). A 15% price rise → 31.5% demand drop (2.1×15%). Price up 15%, volume down 31.5% → net revenue loss. Same pricing action, opposite revenue outcomes — elasticity determines the right strategy.
Q5 — Path analysis (IBM AMOS) reveals that training hours affect customer satisfaction almost entirely through the mediating variable of service quality (indirect β = 0.41) with almost no direct effect (direct β = 0.09, p = 0.31). What does this mean for management decisions?
A
Training does not matter — only service quality drives satisfaction
B
Service quality does not matter — training is the real driver
C
Training works through service quality — so training investment must be paired with service quality measurement to verify the mechanism is functioning
D
The model is wrong — training should affect satisfaction directly
Full mediation: training affects satisfaction primarily through service quality, not directly. This has a critical management implication: if training is delivered but service quality does not improve (perhaps because the training is theoretical rather than practical), satisfaction will not improve either. The mechanism must be monitored. Adding a service quality measurement checkpoint between training completion and satisfaction measurement will tell you whether the mediation pathway is functioning as expected.