Checking your access…
Class · October 25, 2025
A recording of Michael Hlinka's live CFA® Level II class on October 25, 2025 (2 h 3 min) from TA1 Quantitative Methods. It is lesson 11 of 11 in the section, which also has notes (41 pp). It opens free with an account and a short questionnaire.
With the lesson
What this lesson covers
Different formats for the same field, such as dates, are non-uniformity; contradictory data, such as interest expense with zero debt, is inconsistency.
Small training error but much larger cross-validation error signals overfitting and high variance; lasso regularization can prevent overfitting in logistic regression.
When the main concern is the cost of type I errors, the most important evaluation metric is precision.
When comparing logistic regression models on log likelihood, favor the higher number, which is the less negative one.
Covariance stationary needs a constant, finite expected value, constant, finite variance, and constant, finite covariance with itself over a fixed number of periods.
Drafted from the lesson, for Michael to review.