🛠️ Sample size calculator script

💡 Use Case

• The A-B test sample size sheet rewritten as a short script, so that the four assumptions sit in one place, at the top, with their reasons written next to them.
• Written two weeks after the triage test was designed, when I found I had changed the power on one tab of the sheet and not on the other.
• It implements 📐 Sample size for comparing two proportions and prints a small table: the sample per group for the difference I care about, and for half and double that difference.
• The half and double rows are the lesson of 🎓 Lecture 9 - A-B tests and their traps: halve the effect you hope for and the sample you need roughly quadruples.

⚖️ Pros & Cons

Pros

• Assumptions first. Anyone can read lines 1 to 8 and disagree with me before any arithmetic happens.
• The three-row table ends the conversation that starts can we not just run it for a week.
• It is saved with each test, so the sizing is part of the record.

Cons

• It assumes every dispatcher is independent of every other. Dispatchers share an office and a list.
• It answers how many, not how long. How long depends on how many dispatchers a customer has, and for some customers the answer is never.

🔄 Alternatives

• The spreadsheet, for anyone who will not open a script.
• An online calculator. Fine for the number, useless as a record of what was assumed.