maddyhelps

Math 9 · Probability and statistics

Who you ask, and what it predicts

A survey result is only as good as the question asked and the people who answered it. This unit is about the choices that make data trustworthy, and about how probability — in weather forecasts, insurance, sport and medicine — turns past data into a prediction.

The words, first

The idea: Population and sample are the pair everything else hangs on: who you want to know about, and who you actually asked.

WordWhat it means
PopulationThe whole group you want to know about: every student in the school, every voter in Alberta.
SampleThe part of the population you actually ask or measure. Asking everyone is a census.
BiasAnything that pushes results away from the truth in one direction: a leading question, a lopsided sample, bad timing.
Representative sampleA sample whose makeup matches the population's, so its results can stand in for everyone's.
RandomEvery member has the same chance of being chosen.
ProbabilityHow likely an event is, from 0 (impossible) to 1 (certain), as a fraction, decimal or percent.
Theoretical probabilityWorked out from equally likely outcomes. On a fair die, rolling a 5 has probability 1/6.
Experimental probabilityWorked out from what actually happened: successes ÷ trials.
AssumptionSomething taken as true without being checked. For a prediction, usually that the future will behave like the past.

Issues in collecting data

The idea: Seven things can spoil data before any maths is done. Most of them are about the question and the moment, not the numbers.

IssueExample
BiasAsking whether the town needs a skatepark, only at the skatepark
Use of language“Don't you agree the ugly old gym should be replaced?”
EthicsCollecting answers “for a class project” and then passing them to an advertiser
CostTesting how long every light bulb in a factory lasts — none would be left to sell
Time and timingAsking about favourite seasons during a January cold snap
PrivacyA named form asking about your family's income
Cultural sensitivityAsking the whole school which gift they got for Christmas

Worked example: fixing a question. “Don't you agree that the school should stop wasting money on new uniforms?” has two problems. “Don't you agree” invites a yes, and “wasting” has already decided the answer. A neutral version: “Should the school buy new uniforms this year? Yes / No / No opinion.” It names the issue, takes no side, and lets people who do not care say so.

Populations and samples

The idea: A sample is only useful if it looks like the population. Choosing at random is how you get there without having to guess what matters.

A school of 600 students wants to know how students get to school. Asking 60 of them is a sample. Six ways to choose the 60:

MethodHow the 60 are chosenFair?
Simple randomA computer picks 60 names from all 600Yes
SystematicEvery 10th name on the list, from a random startYes, if the list has no hidden pattern
StratifiedA random 10% of each grade: 18 of 180 Grade 7s, 20 of 200 Grade 8s, 22 of 220 Grade 9sYes, with every grade fairly represented
ClusterTwo homerooms of 30 picked at random, everyone in them askedUsually, if homerooms are alike
ConvenienceThe first 60 through the front doorsNo: bus riders may use another door
Voluntary responseA poll on the school websiteNo: mostly people with strong feelings reply

Worked example: from sample to population. In a fair sample of 60, 18 students ride the bus. That is 18 ÷ 60 = 0.30, so the prediction for the whole school is 0.30 × 600 = 180 bus riders. The prediction assumes the sample is representative. Had the 60 been the first through the front doors, it would undercount bus riders, and every number built on it would be wrong in the same direction.

Probability in society

The idea: A probability is a prediction built from data and assumptions. Knowing the assumption tells you how far to trust it.

Theoretical and experimental. On a fair die, P(5 or 6) = 2/6 = 1/3. Roll it 60 times and get a 5 or a 6 on 23 of them, and the experimental probability is 23/60 ≈ 0.38. The two differ because 60 rolls is not many; with more rolls, the experimental value tends to settle closer to 1/3.

Worked example. A basketball player has made 34 of her last 40 free throws. Predict how many of her next 20 she will make.

  1. Experimental probability: 34 ÷ 40 = 0.85, or 85%.
  2. Prediction: 0.85 × 20 = 17 free throws.
  3. The assumptions: she keeps shooting as she has been — no injury, no extra pressure. Change that and the prediction weakens.

Where you meet it.

  • Weather. A 40% chance of rain means that on days forecast like this one, measurable rain falls at a given spot about 4 times in 10 — not rain for 40% of the day.
  • Insurance. Prices are set from claims data. New drivers have more collisions on average, so they usually pay more: a prediction about a group, applied to each person in it.
  • Sport. A goalie's save percentage of .915 is an experimental probability: 915 saves for every 1000 shots.
  • Medicine. A side effect in 2 of every 100 patients in trials predicts your chance only as far as you are like the people tested.

A probability is not the whole decision. A 30% chance of rain might mean no umbrella for the walk to school, but a rented tent for a wedding. What is at stake decides.

What costs marks

The idea: Most of these are naming the wrong method, or not naming the problem at all.

  • Calling a convenience sample random. Easy to reach is not the same as random.
  • Mixing up voluntary response and convenience. Voluntary means people chose themselves; convenience means you chose whoever was handy.
  • Saying a question is biased without saying why. Name the words that lead.
  • Treating an experimental probability as exact. It is an estimate, and it improves with more trials.
  • Forgetting the assumption when asked how reliable a prediction is.

Practice this unit Build a quiz