STUDY GUIDE · 3 INTERACTIVE LABS · ABOUT 12 MIN

When a billionaire walks into a bar, the average customer becomes a millionaire.

Nobody in the bar got richer, but the mean says otherwise. Statistics is the craft of summarising data honestly, and of knowing how much to trust a conclusion drawn from a sample. Every IQ score, including ours, is built on it.

THE NORMAL CURVE, IN IQ POINTS
100 ± 1568.3%
100 ± 3095.4%
100 ± 4599.7%
above 1302.3%
IQ tests are scaled to a mean of 100 and a standard deviation of 15. The share of people within 1, 2 and 3 SDs follows the 68-95-99.7 rule.

What is statistics?

Statistics is the science of collecting, summarising and interpreting data. Descriptive statistics summarise what you have (averages, spread, charts). Inferential statistics use a sample to draw conclusions about a whole population, with a measured amount of uncertainty.

It is how drugs are approved, polls are reported, experiments are judged and IQ scores are standardised. Done badly, it's also how misleading headlines are made.

MeanAdd all values, divide by how many.
MedianThe middle value when sorted.
Standard deviationTypical distance of values from the mean.
p-valueHow surprising the data would be if there were no real effect.
LAB 1 · MEAN, MEDIAN AND SPREAD

Which average tells the truth?

The mean adds everything up and divides; the median is the middle value. With symmetric data they agree. Add one extreme value and watch which one moves.

£0ksalary, £ thousands£120k
£39.1kmean
£37kmedian
£9.9kstandard deviation
12values
Mean and median are close but not equal: salaries are slightly skewed by the higher earners. Now let the billionaire in.
meanmedianmean ± 1 standard deviation
LAB 2 · CORRELATION

Dial the correlation, see the cloud

The correlation coefficient r runs from −1 (perfect downhill line) through 0 (no linear pattern) to +1 (perfect uphill line). r² is the share of variation in y that the line explains.

+0.56measured r (60 points)
32%r² · variation explained
Moderate: a visible trend, but plenty of exceptions. Most real-world links between, say, test scores and later outcomes look like this.
CORRELATION ≠ CAUSATIONIce-cream sales and drownings rise together, not because one causes the other, but because both rise in hot weather. A hidden third factor (a confounder) can create a strong correlation between unrelated things.
LAB 3 · SAMPLING AND CONFIDENCE INTERVALS

What does "95% confident" actually mean?

We know the true average IQ is 100. Draw 20 random samples, and for each compute a 95% confidence interval. About 19 in 20 should catch the true value: that's the 95%. Bigger samples give narrower intervals.

IQ 85true mean 100IQ 115
20 / 20intervals contain 100
± 5.9margin of error
All 20 caught the true mean this time; draw again and one will usually miss. Quadrupling the sample size halves the margin of error, because it shrinks with √n.95% CI = x̄ ± 1.96 × 15 / √n
catches the true meanmisses it

From plague records to p-values

1662John Graunt.Analysed London's weekly death records, founding demography and statistics.
1809The normal curve.Gauss used it to describe errors in astronomical measurements.
1854John Snow's map.Plotting cholera deaths around a water pump helped show the disease was waterborne.
1908Student's t-test.William Gosset, working for Guinness, published under the pen name 'Student'.
1925Fisher's methods.Ronald Fisher popularised p-values, significance testing and randomised experiments.
2016The ASA statement.The American Statistical Association warned against treating p < 0.05 as proof.

Further reading

The Art of StatisticsDavid Spiegelhalter · learning from data, clearly explained
How to Lie with StatisticsDarrell Huff · the 1954 classic on misleading numbers
Naked StatisticsCharles Wheelan · the important ideas without the dread
Test yourself: 20 statistics questionsProbability guide

Now see where you fall on the curve. Take the test.

Our full IQ test is scaled to mean 100, SD 15, just as above: 60 questions, about 30 minutes.

Take the full IQ test