Assess techniques like ridge, elastic net, PCA, and channel aggregation that stabilise models when predictors are highly correlated.
Multicollinearity inflates the variance of ______ estimates.
impression
creative
coefficient
bounce
When predictors correlate highly, standard errors grow, making elasticities unstable and difficult to interpret.
Variance Inflation Factor (VIF) above ______ signals problematic collinearity.
5‑10
1
100
0.1
Many practitioners flag a VIF over 5 as moderate concern and above 10 as severe multicollinearity.
Ridge regression mitigates multicollinearity by adding a penalty on ______ magnitude.
cookie
coefficient
error
spend
L2 penalty shrinks correlated coefficients toward zero, trading bias for lower variance.
Elastic Net combines L1 and L2 penalties, enabling both shrinkage and ______.
creative rotation
feature selection
CPC inflation
impression share
The L1 component can zero out redundant predictors, simplifying the model.
Principal Component Analysis addresses collinearity by creating ______ predictors.
orthogonal
duplicate
scaled
segmented
PCA transforms correlated variables into uncorrelated principal components, though interpretability can suffer.
Aggregating highly correlated channels into a single meta‑channel reduces collinearity but sacrifices ______ granularity.
time‑series
creative
device
channel‑level
Grouping simplifies the model but hides individual performance differences.
Stepwise variable selection can exacerbate collinearity by repeatedly choosing variables based on ______.
marginal fit
creative freshness
pixel load
CTR
Adding variables purely on marginal R² can latch onto noise shared among correlated predictors.
Centering variables (subtracting mean) does ______ to the correlation matrix.
invert
nothing
increase
zero out
Centering eases numerical stability but does not change correlation; collinearity remains.
High correlation among spend series often arises from simultaneous ______.
creative swaps
domain changes
pixel outages
campaign bursts
Marketers tend to scale budgets across channels at the same time, inflating cross‑channel correlation.
The 2025 MMM Guide suggests using elastic net with cross‑validated penalty alpha to balance bias and ______.
creative score
cookie match
variance
reach
Optimally tuned penalties achieve stable yet interpretable coefficients.
Starter
Review the basics.
Solid
Nice work—refine the details.
Expert!
Exceptional command of the topic.
