Mathematics
Confidence Intervals
Quick fact
A 95% confidence interval does NOT mean there is a 95% chance the true value lies within that specific interval. Instead, it means that if you repeated the sampling process many times, about 95% of the calculated intervals would capture the true population parameter.
Why this is interesting
You've seen polls with "plus or minus 3 percent" — but what does that really tell you? And why do statisticians say they are 95% confident?
Read the full explanation
Understanding Confidence Intervals
Imagine you are fishing in a lake with a net. The net represents your sample, and the fish you catch is your sample statistic. But the true big fish (the population parameter) is somewhere else in the lake. A confidence interval is like extending your net to a larger area around where you fished. Different samples (different casts) will produce different interval lengths and positions. The confidence level (e.g., 95%) tells you that over many casts, roughly 95 out of 100 of those extended nets will actually contain the true fish. So a confidence interval is a way of saying, "Based on this sample, we believe the true value is within this range, and we have a certain level of trust in this method."
A deeper explanation
Confidence intervals rely on the concept of a sampling distribution. For a given statistic like the mean, its sampling distribution is approximately normal (by the Central Limit Theorem) with a standard error that measures variability. The interval is constructed by taking the sample estimate and adding/subtracting a margin of error: margin = critical value × standard error. The critical value comes from the normal or t-distribution, chosen to achieve the desired confidence level (e.g., 1.96 for 95% confidence in a normal distribution). The interval's width reflects uncertainty: larger samples reduce standard error, producing narrower intervals. This mechanism connects sample precision to population inference, making confidence intervals essential for evidence-based reasoning. Applications range from clinical trials to election forecasts, where decisions hinge on whether an observed effect is credible.