Create intermediate Works with: Any AI

Missing Data Strategy Interview

Interview first Has example

Decide how to handle nulls case by case.

The Open buttons send the prompt text as currently shown below to that site only when you click. Fill in the placeholders first.

Interview me before recommending how to handle missing values. Ask one question at a time and wait: (1) the dataset and what each column means; (2) how much is missing per column (I can paste df.isna().mean()); (3) why values might be missing (not collected, not applicable, error, refused); (4) whether missing is informative (for example blank discount means no discount); (5) what the data will be used for (reporting, a model, a join); (6) how costly a wrong guess is. Starting description: [DATASET_DESCRIPTION].

Then produce a table: Column | Missing % | Likely mechanism (completely at random, at random given other columns, not at random) | Options (drop rows, drop column, fill with constant, fill with group median, forward fill, model-based, leave as a category) | Recommended option and why | Code | Side effect to check.
Add a missing-indicator rule (when to add a was_missing column), a warning about filling before splitting data for a model (leakage), and a before-and-after check of the distribution of each filled column. Be explicit where the right answer depends on a business decision I must make.

Fill in before use

Optional: type your details and the prompt above updates. Nothing is saved or sent.

Replace each bracketed placeholder with your own details, then paste into Any AI or another assistant.

Q1: Which columns have missing values, and how many? A: income 18%, age 3%.
Q2: Why might they be missing? A: income is optional on the form.
Reflection: missing not at random is likely for income.
Q3: What will the data be used for? A: a model predicting loan approval.
Recommendation
Age (3%): median imputation plus a missing flag.
Income (18%): keep a missing flag; compare complete-case results with an imputed version.
Do not drop rows blindly.
Document the choice and its risk: imputed values may understate variation.
Sensitivity check: rerun the model with and without imputation and compare metrics.

• missing-data