Module 02 Summary

What this module established

Five decisions define the analysis: outcome, predictor, grouping, unit and time window. The unit is the one that catches people — 8 rows from 3 drivers gives n = 8 or n = 3, and every standard error divides by the square root of it.

Carry forward

• pandas skips NaN in almost every aggregation without saying so. Put count in every summary, because it reports what was actually used.
• duplicated() finds rows identical in every column, which repeated measurements are not. Compare len(df) with df[unit].nunique() instead.
• statsmodels takes the alphabetically first level as the reference. economy sorts before standard, so a delivery model silently compares against the slowest plan.
• Run the check dictionary before the test. A duplicated id or a group of three will not stop ttest_ind returning a confident answer.

Before moving on

Move on when you can state what one row is, how many distinct units you have, and what your reference level is.

Mark complete