When estimating a population mean, the population standard deviation is usually unknown. Replacing it with the sample standard deviation adds uncertainty. The t-distribution accounts for that extra uncertainty, especially in small samples.
The standard error uses sample spread
For a sample of quantitative observations, estimate the population mean with The estimated standard error is where is the sample standard deviation.
A one-sample interval is
with degrees of freedom. The critical value depends on both confidence level and degrees of freedom. A t-distribution is symmetric about zero with heavier tails than the standard normal; those tails reflect uncertainty from estimating the population spread.
As degrees of freedom increase, the t-distribution approaches the standard normal. Using 1.96 automatically for every 95% mean interval can make small-sample intervals too narrow.
Check the design and shape conditions
The sample should be random or otherwise support the stated inference. Observations should be independent; for sampling without replacement, check the sample is at most about 10% of the population.
For small samples, the population should be approximately normal, with no strong skewness or outliers evident in the sample. Larger samples make the procedure more robust to moderate nonnormality, but severe skewness or influential outliers still deserve attention. A numerical sample-size threshold is a guideline, not a substitute for examining the data.
The data must be quantitative. A binary success proportion has a different sampling model and should not be substituted mechanically into a mean-interval procedure in this introductory setting.
Example: construct a 95% interval
A random sample of 16 batteries has mean life 42 hours and standard deviation 8 hours. Suppose a plot shows no strong skewness or outliers, and independence is reasonable. The standard error is hours.
With 15 degrees of freedom, the 95% critical value is about 2.131. The margin of error is giving an interval approximately hours.
We are 95% confident that the population mean battery life lies between about 37.7 and 46.3 hours. This does not say that 95% of individual batteries last within that range. Individual lifetimes vary more widely than sample means.
Why degrees of freedom are minus one
Once the sample mean is fixed, the deviations from it sum to zero. Knowing minus one deviations determines the final deviation. This constraint is reflected in the sample-variance estimate and the reference distribution's degrees of freedom.
You do not subtract another degree of freedom for constructing an interval. The usual one-sample mean procedure uses minus one, regardless of the confidence level.
Example: compare precision across sample sizes
If the same sample standard deviation 8 occurred with 64 observations, the standard error would be 1 hour. At 63 degrees of freedom, the 95% critical value is about 1.998, giving a margin near 2 hours.
The larger sample reduces uncertainty both by dividing by a larger square root and by using a critical value closer to the normal value. The sample standard deviation describes individual spread; the standard error describes uncertainty in the estimated mean. Those are different quantities.
Increasing confidence from 95% to 99% would increase the critical value and widen the interval for the same data. More confidence requires a more conservative range, not a more precise estimate.
Interpret confidence and limitations honestly
Across repeated random samples, a 95% interval procedure captures the fixed population mean about 95% of the time under its assumptions. It does not correct biased recruitment, inaccurate measurements, or a mismatched target population.
A interval estimates a population mean, not the next individual observation. Predicting a future battery's life requires a different, usually wider, interval because it includes individual variability as well as uncertainty in the mean.
Try it yourself
A sample has equals 25, mean 100, and standard deviation 15. Using find the 95% interval.
Answer: The standard error is The margin is so the interval is approximately The degrees of freedom are 24.
What to remember
Use when population spread is estimated by , check independence and distribution shape, and distinguish the spread of individuals from the uncertainty of the sample mean.