DA0-002 Data Acquisition and Preparation Practice Question

Answer analysis

Option-by-option breakdown

For each option: why learners choose it and why it is or isn't the right answer here.

• ✗

  df['age'].replace(np.nan, df['age'].mean())

  Why it's wrong here

  Using mean() computes the arithmetic average, not the median, so the imputed values are wrong for skewed age data. replace() itself is tempting because it targets specific values, and would suit substituting a known sentinel like -1 or 'N/A' rather than statistically derived imputation.

• ✗

  df['age'].dropna()

  Why it's wrong here

  `dropna()` deletes rows containing missing values, so it removes records rather than imputing the median — the 'age' column would lose data and no replacement occurs. It is tempting because `dropna()` is the standard pandas tool for discarding incomplete rows during exploratory cleaning, and would be correct if the requirement were to exclude incomplete records entirely rather than fill them.

• ✓

  df['age'].fillna(df['age'].median())

  Why this is correct

  fillna substitutes missing entries, and passing the column's median supplies a single imputation value computed across non-null ages. This directly satisfies the requirement to replace NaN values in 'age' with the median, preserving row count without dropping records.

• ✗

  df['age'].interpolate()

  Why it's wrong here

  interpolate() fills gaps by linear estimation between neighbouring index points, producing values that depend on row order rather than the column's central tendency, so it never computes a median. It is tempting for time-series or ordered numeric data where smooth continuity between points is the desired behaviour.