The missing-value fork: impute, exclude, or escalate?
Navigate a missing-data decision without silently biasing the result.
The fork Rosa’s conversion_rate model has 18% missing industry codes, and 70% of those missing rows are from one acquisition channel. Deleting them would make the script clean and the story false. The choice changes what future-you can verify. CRISP-DM CRISP-DM’s data-understanding phase exists to prevent modeling decisions from outrunning data meaning. Missing values are not just blanks; they can be random noise, collection failure, business process signal, or a join defect. The mechanism is bias control. If missingness is related to the outcome or a key segment, deletion changes the population. In R, you can profile missingness with grouped…
Sign up free — one personalized lesson every day, matched to your role and goals.
Already have an account? Sign in