1. Read a confidence interval

You are given the interval. You interpret it, check it, and say what changes.

The rules

Interpretation sentence. We are [level]% confident that the true [mean / proportion] of [what was measured] for [the whole population] is between [L] and [U] [units].
Validity sentence for a mean. The sample is random, and n = [n] is at least 30, so by the Central Limit Theorem the sample mean is approximately normal. The interval is valid.
If n is under 30: it is valid only if the population itself is approximately normal, and a small sample cannot confirm that.
Validity sentence for a proportion. The sample is random, and it has [x] successes and [n − x] failures. Both are at least 5, so the normal approximation holds. The interval is valid. The class rule is 5. The textbook uses 8 for intervals. State the two counts and you are covered either way.

What makes an interval wider or narrower

ChangeEffect on the intervalWhy
Higher confidence (95% to 99%)Wider, so less preciseThe multiplier grows. To be surer of catching the true value you need a wider net.
Lower confidence (95% to 90%)NarrowerThe multiplier shrinks.
Larger sampleNarrower, so more preciseThe standard error s/√n shrinks.
More spread in the data (larger s)WiderThe standard error grows.
4 times the sampleHalf the margin of error√4 = 2 sits in the denominator.
Precise means narrow. Raising the confidence level makes the interval less precise. The way to get more precision at the same confidence is a larger sample.

Worked class example: olive oil

A random sample of 81 bottles has a mean of 473 ml and a standard deviation of 3 ml. The 95% confidence interval for the mean fill is (472.34, 473.66) ml.

Interpret the interval.class example

Answer

We are 95% confident that the true mean amount of olive oil per bottle, for all bottles this machine fills, is between 472.34 ml and 473.66 ml.

Were the assumptions for this method met?review class

Answer

Yes. The 81 bottles are a random sample. The sample size is 81, which is at least 30, so by the Central Limit Theorem the sample mean is approximately normal even if individual fills are not. The population standard deviation is unknown, so a t interval with 80 degrees of freedom is the right method.

Your boss says: "I want this more precise. Use 99% instead." Respond without recalculating.review class

Answer

A 99% interval would be wider, so it would be less precise, not more. The mean, the standard deviation and the sample size stay the same. Only the multiplier changes, and it gets larger (about 2.64 instead of 1.99), so the margin of error grows. To be more confident that the interval catches the true mean, it has to cover more values. If the boss wants a narrower interval at 95%, the fix is a larger sample.

From the interval alone, find the sample mean and the margin of error.

Answer

The sample mean is the midpoint: (472.34 + 473.66)/2 = 473 ml. The margin of error is half the width: (473.66 − 472.34)/2 = 0.66 ml.

True or false: there is a 95% probability that the true mean fill is between 472.34 and 473.66 ml.Homework 9, Q1C

Answer

False. The true mean is a fixed number, so it is either in this interval or it is not. The 95% describes the method: about 95% of the intervals built this way from repeated samples would contain the true mean.

A co-worker says: "95% of all bottles contain between 472.34 and 473.66 ml." Why is that wrong?Homework 9, Q3D

Answer

The interval estimates the population mean. It does not describe individual bottles. Individual fills vary with a standard deviation of about 3 ml, so most single bottles fall well outside a range that is only 1.3 ml wide.

Why does this interval use t and not z?

Answer

The population standard deviation σ is unknown, so it is replaced by the sample standard deviation s. That adds uncertainty, and the t distribution with n − 1 = 80 degrees of freedom accounts for it. With 80 degrees of freedom the t multiplier (1.990) is almost the same as the z multiplier (1.96).

More practice

A 90% interval for the mean histamine content of worker bees, from a sample of 7 bees, is (490.0, 786.3) nanograms. Is the interval statistically valid?Homework 9, Q1B

Answer

Only if histamine content is approximately normally distributed in the population. The sample has 7 bees, which is far below 30, so the Central Limit Theorem does not apply. With so few values the normality assumption cannot really be checked, so the validity is in doubt.

An 80% interval for the mean breaking strength of cotton thread, from 50 measurements, is (206.7, 213.3) grams. Without recalculating, would a 90% interval be wider or narrower?Homework 9, Q3C

Answer

Wider. The sample mean, standard deviation and sample size are unchanged. The multiplier for 90% is larger than the one for 80%, so the margin of error is larger.

A random sample of 500 airline customers has 40 who qualify for a club. The 95% interval for the proportion who qualify is (0.056, 0.104). Interpret it and say whether it is valid.Homework 9, Q5

Answer

We are 95% confident that the true proportion of all the airline's current customers who would qualify is between 0.056 and 0.104, that is between 5.6% and 10.4%.

It is valid. The sample is random, and it contains 40 successes and 460 failures, both far above the minimum of 5.

Two analysts use the same data. One reports (20, 30) and the other reports (22, 28). Which used the higher confidence level?

Answer

The one who reported (20, 30). Both intervals are centred at 25 because the data are the same. The wider interval has the larger multiplier, which means the higher confidence level.

A 95% interval for a mean is (40, 60) from a sample of 25. About what would the interval be with a sample of 100 and the same mean and standard deviation?

Answer

The margin of error is 10. Four times the sample size cuts the standard error in half, so the margin becomes about 5 and the interval is about (45, 55). The multiplier also shrinks a little because the degrees of freedom rise, so the true answer is slightly narrower still.