Polling & Survey Sampling: Margin of Error, Weighting & Poll Averages

What you'll build

What you will be able to do

Each lab ends with a concrete result that you build yourself and the checker confirms, cell by cell.

• Put a 95 % margin of error on every share in a referendum poll, build the confidence interval and find the single margin a pollster prints under the headline.
• Work out how many completed interviews, invitations and dollars the next poll needs for margins from ±5 to ±2 points, and what it costs to report every district.
• See how much the margin shrinks when a poll reaches a large part of a small ward, and plan a new island poll with the corrected sample size.
• Split 1,200 interviews over four districts in two ways and compare the district margins with the margin for the whole city.
• Weight six age × area cells of a referendum poll so each cell counts as much as it does in the electorate, and measure how much the Yes share moves.
• Measure how many interviews the weighted poll is really worth, widen the margin accordingly, and estimate the design effect of a clustered door-to-door poll.
• Combine eight polls into one average that trusts big and recent polls more, compare it with the simple average and chart the trend.
• Calculate the margin of error of a lead in two polls, decide whether each lead is clear and put a probability on Yes being ahead.