Question 29
Domain 3: Model DevelopmentWhich symptom most strongly suggests excessive variance?
Correct answer: A
Explanation
Excessive variance means the model fits the training data too closely and does not generalize well. That is shown by "low training error and much worse validation error," because the gap between training and validation performance signals overfitting.
Why each option is right or wrong
A. Low training error and much worse validation error
Excessive variance is diagnosed by a large generalization gap: the model performs very well on the data it was fit on but degrades substantially on unseen data. In this question, the combination of very low training error with much higher validation error is the classic pattern of overfitting, indicating the model has captured noise or idiosyncrasies in the training set rather than a stable underlying relationship.
B. High training error and equally high validation error
High training and validation error indicates high bias or underfitting, not variance.
C. Perfectly matched training and validation performance at a low score
Matched low scores suggest consistent performance, but not the train-validation gap of variance.
D. A missing label column
Missing label columns are a data quality issue, not a bias-variance performance symptom.