The normal distribution — Data, 14–17 years
A model for measurements that gather around a typical value, with fewer results as you move away from it.
Idea
Many measurements form a mound-shaped pattern: most values sit near the centre, while very small and very large values are uncommon. Heights, repeated measurement errors and some test scores can look like this, although real data are never perfectly smooth.
Why use it?
Researchers needed a simple way to describe variation caused by many small influences acting together. The normal model helps estimate what is common and what is unusual, but it is an approximation, not a promise that every dataset follows the same shape.
Worked example
Suppose reaction times are modelled as normal with mean 250 ms and standard deviation 30 ms. About 68% lie from 220 to 280 ms, because that is one standard deviation on either side. About 95% lie from 190 to 310 ms, using two standard deviations.
Common trap
A bell shape can look convincing, so it is tempting to force every dataset into a normal model. That is reasonable because the model is powerful and familiar, but it fails for strongly skewed data, bounded scores or data with several distinct groups.
Where it appears
Quality control uses this model to flag products whose measurements are unusually far from the target. Medical reference ranges and exam-score analysis may also use it, but professionals check whether the data fit before relying on its percentages.
Keep exploring
Other languages
Loading MyLeoNes™…