Online coding and Python classes in Ashford, Kent

The learner pulls Census 2021 table TS017, household size, from the Nomis API for Ashford and for England. Taking Ashford's counts away from England's leaves the rest of the country, a comparison group we built ourselves. Households with five or more people are merged into one group so that every group has plenty of cases.

A chi-square goodness-of-fit test asks whether Ashford's 53,582 households could plausibly have been drawn from the rest of England's mix. The answer is a resounding no: the test statistic is 498 on four degrees of freedom, with a p-value of about 1.7 times ten to the power minus 106. Yet the table shows the gap is not dramatic. Ashford has fewer people living alone and a few more four-person homes. Two measures put a size on it. Cohen's w comes out at 0.096, just under 0.1, the value conventionally labelled a "small" effect. The total variation distance says only 4.2% of Ashford's households would have to change group for the two mixes to match exactly.

This is statistical power in action. The real difference never changes, but with 100 households the test spots it in barely one sample in nine, and with 3,000 it almost never misses. Meanwhile the effect size settles close to 0.1 once samples reach about a thousand. The small samples show a second trap: random noise pushes their typical w upwards, so a small study can both miss a real effect and exaggerate the one it sees.

Reading statistics with that care lets Ashford teenagers question AI-written reports instead of repeating them, one more reason to learn Python in 2026. The longer argument is in why teenagers should still learn to code in 2026.