In statistics, a confidence interval (CI) is an interval estimate of a population parameter. Instead of estimating the parameter by a single value, an interval likely to include the parameter is given. Thus, confidence intervals are used to indicate the reliability of an estimate. How likely the interval is to contain the parameter is determined by the confidence level or confidence coefficient. Increasing the desired confidence level will widen the confidence interval.
For example, a CI can be used to describe how reliable survey results are. In a poll of election voting-intentions, the result might be that 40% of respondents intend to vote for a certain party. A 95% confidence interval for the proportion in the whole population having the same intention on the survey date might be 36% to 44%. All other things being equal, a survey result with a small CI is more reliable than a result with a large CI and one of the main things controlling this width in the case of population surveys is the size of the sample questioned. Confidence intervals and interval estimates more generally have applications across the whole range of quantitative studies.
In the above, the 95% associated with the confidence interval is called the confidence level of the interval: this is defined formally below.
For a given proportion p (where p is the confidence level), a confidence interval for a population parameter is an interval that is calculated from a random sample of an underlying population such that, if the sampling was repeated numerous times and the confidence interval recalculated from each sample according to the same method, a proportion p of the confidence intervals would contain the population parameter in question. In unusual cases, a confidence set may consist of a collection of several separate intervals, which may include semi-infinite intervals, and it is possible that an outcome of a confidence-interval calculation could be the set of all values from minus infinity to plus infinity.
Confidence intervals are the most prevalent form of interval estimation. Interval estimates may be contrasted with point estimates and have the advantage over these as summaries of a dataset in that they convey more information – not just a "best estimate" of a parameter but an indication of the precision with which the parameter is known.
Confidence intervals play a similar role in frequentist statistics to the credibility interval in Bayesian statistics. However, confidence intervals and credibility intervals are not only mathematically different; they have radically different interpretations.
Confidence regions generalise the confidence interval concept to deal with multiple quantities. Such regions can indicate not only the extent of likely estimation errors but can also reveal whether (for example) the estimate for one quantity is too large then the other is also likely to be too large. See also confidence bands.
In applied practice, confidence intervals are typically stated at the 95% confidence level. However, when presented graphically, confidence intervals can show several confidence levels, for example 50%, 95% and 99%.
Confidence intervals are constructed on the basis of a given dataset: x denotes the set of observations in the dataset, and X is used when considering the outcomes that might have been observed from the same population, where X is treated as a random variable whose observed outcome is X = x. A confidence interval is specified by a pair of functions u(.) and v(.) and the confidence interval for the given data set is defined as the interval (u(x), v(x)). To complete the definition of a confidence interval, there needs to be a clear understanding of the quantity for which the CI provides an interval estimate. Suppose this quantity is w. The property of the rules u(.) and v(.) that makes the interval (u(x),v(x)) closest to what a confidence interval for w would be, relates to the properties of the set of random intervals given by (u(X),v(X)): that is treating the end-points as random variables. This property is the coverage probability or the probability c that the random interval includes w,
Here the endpoints U = u(X) and V = v(X) are statistics (i.e., observable random variables) which are derived from values in the dataset. The random interval is (U, V).
For the above to provide a viable means to statistical inference, something further is required: a tie between the quantity being estimated and the probability distribution of the outcome X. Suppose that this probability distribution is characterised by the unobservable parameter θ, which is a quantity to be estimated, and by other unobservable parameters φ which are not of immediate interest. These other quantities φ in which there is no immediate interest are called nuisance parameters, as statistical theory still needs to find some way to deal with them.
The definition of a confidence interval for θ is, for a given α,
The number (sometimes reported as a percentage (100%·) is called the confidence level or confidence coefficient. Most standard books adopt this convention, where α will be a small number. Here is used to indicate the probability when the random variable X has the distribution characterised by . An important part of this specification is that the random interval (U, V) covers the unknown value θ with a high probability no matter what the true value of θ actually is.
Note that here need not refer to an explicitly given parameterised family of distributions, although it often does. Just as the random variable X notionally corresponds to other possible realisations of x from the same population or from the same version of reality, the parameters indicate that we need to consider other versions of reality in which the distribution of X might have different characteristics.
Here is used to indicate the probability over the joint distribution of the random variables (X,Y) when this is characterised by parameters .
For non-standard applications it is sometimes not possible to find rules for constructing confidence intervals that have exactly the required properties. But practically useful intervals can still be found. The coverage probability for a random interval is defined by
and the rule for constructing the interval may be accepted as providing a confidence interval if
to an acceptable level of approximation.
A Bayesian interval estimate is called a credible interval. Using much of the same notation as above, the definition of a credible interval for the unknown true value of θ is, for a given α,
Here Θ is used to emphasize that the unknown value of is being treated as a random variable. The definitions of the two types of intervals may be compared as follows.
Note that the treatment of the nuisance parameters above is often omitted from discussions comparing confidence and credible intervals but it is markedly different between the two cases.
In some simple standard cases, the intervals produced as confidence and credible intervals from the same data set can be identical. They are always very different if moderate or strong prior information is included in the Bayesian analysis.
When applying fairly standard statistical procedures, there will often be fairly standard ways of constructing confidence intervals. These will have been devised so as to meet certain desirable properties, which will hold given that the assumptions on which the procedure rely are true. In non-standard applications, the same desirable properties would be sought. These desirable properties may be described as: validity, optimality and invariance. Of these "validity" is most important, followed closely by "optimality". "Invariance" may be considered as a property of the method of derivation of a confidence interval rather than of the rule for constructing the interval.
For non-standard applications, there are several routes that might be taken to derive a rule for the construction of confidence intervals. Established rules for standard procedures might be justified or explained via several of these routes. Typically a rule for constructing confidence intervals is closely tied to a particular way of finding a point estimate of the quantity being considered.Sample statistics: This is closely related to the method of moments for estimation. A simple example arises where the quantity to be estimated is the mean, in which case a natural estimate is the sample mean. The usual arguments indicate that the sample variance can be used to estimate the variance of the sample mean. A naive confidence interval for the true mean can be constructed centered on the sample mean with a width which is a multiple of the square root of the sample variance.Likelihood theory: Where estimates are constructed using the maximum likelihood principle, the theory for this provides two ways of constructing confidence intervals or confidence regions for the estimates.Estimating equations: The estimation approach here can be considered as both a generalization of the method of moments and a generalization of the maximum likelihood approach. There are corresponding generalizations of the results of maximum likelihood theory that allow confidence intervals to be constructed based on estimates derived from estimating equations.Via significance testing: If significance tests are available for general values of a parameter, then confidence intervals/regions can be constructed by including in the 100p% confidence region all those points for which the significance test of the null hypothesis that the true value is the given value is not rejected at a significance level of (1-p).
To get an impression of the expectation μ, it is sufficient to give an estimate. The appropriate estimator is the sample mean:
The sample shows actual weights , with mean:
If we take another sample of 25 cups, we could easily expect to find values like 250.4 or 251.1 grams. A sample mean value of 280 grams however would be extremely rare if the mean content of the cups is in fact close to 250g. There is a whole interval around the observed value 250.2 of the sample mean within which, if the whole population mean actually takes a value in this range, the observed data would not be considered particularly unusual. Such an interval is called a confidence interval for the parameter μ. How do we calculate such an interval? The endpoints of the interval have to be calculated from the sample, so they are statistics, functions of the sample and hence random variables themselves.
In our case we may determine the endpoints by considering that the sample mean from a normally distributed sample is also normally distributed, with the same expectation μ, but with standard error (grams). By standardizing we get a random variable
dependent on μ, but with a standard normal distribution independent of the parameter μ to be estimated. Hence it is possible to find numbers −z and z, independent of μ, where Z lies in between with probability 1 − α, a measure of how confident we want to be. We take 1 − α = 0.95. So we have:
This might be interpreted as: with probability 0.95 to one we will choose a confidence interval in which we will meet the parameter μ between the stochastic endpoints, but that does not mean that possibility of meeting parameter μ in confidence interval is 95% :
Every time the measurements are repeated, there will be another value for the mean of the sample. In 95% of the cases μ will be between the endpoints calculated from this mean, but in 5% of the cases it will not be. The actual confidence interval is calculated by entering the measured weights in the formula. Our 0.95 confidence interval becomes:
This interval has fixed endpoints, where μ might be in between (or not). There is no probability of such an event. We cannot say: "with probability (1 − α) the parameter μ lies in the confidence interval." We only know that by repetition in 100(1 − α) % of the cases μ will be in the calculated interval. In 100α % of the cases however it doesn't. And unfortunately we don't know in which of the cases this happens. That's why we say: "with confidence level 100(1 − α) % μ lies in the confidence interval."
The figure on the right shows 50 realisations of a confidence interval for a given population mean μ. If we randomly choose one realisation, the probability is 95% we end up having chosen an interval that contains the parameter; however we may be unlucky and have picked the wrong one. We'll never know; we're stuck with our interval.
has a Student's t-distribution with n − 1 degrees of freedom. Note that the distribution of T does not depend on the values of the unobservable parameters μ and σ2; i.e., it is a pivotal quantity. If c is the 95th percentile of this distribution, then
(Note: "95th" and "0.9" are correct in the preceding expressions. There is a 5% chance that T will be less than −c and a 5% chance that it will be larger than +c. Thus, the probability that T will be between −c and +c is 90%.)
and we have a theoretical (stochastic) 90% confidence interval for μ.
After observing the sample we find values for and s for S, from which we compute the confidence interval
an interval with fixed numbers as endpoints, of which we can no more say there is a certain probability it contains the parameter μ. Either μ is in this interval or isn't.
For users of frequentist methods, various interpretations of a confidence interval can be given.
In each of the above, the following applies. If the true value of the parameter lies outside the 90% confidence interval once it has been calculated, then an event has occurred which had a probability of 10% (or less) of happening by chance.
Users of Bayesian methods, if they produced an interval estimate, would by contrast want to say "My degree of belief that the parameter is in fact in this interval is 90%" . See Credible interval. Disagreements about these issues are not disagreements about solutions to mathematical problems. Rather they are disagreements about the ways in which mathematics is to be applied.
When a study involves multiple statistical tests, some laymen assume that the confidence associated with individual tests is the confidence one should have in the results of the study itself. In fact, the results of all the statistical tests conducted during a study must be judged as a whole in determining what confidence one may place in the positive links it produces. If a researcher conducting a study performs 40 statistical tests at 95% confidence, she can expect about two of the tests to return false positives. If she in fact finds 3 links, the confidence associated with those links 'as the result of the survey' is actually about 32%; it's what she should expect to see two-thirds of the time.
The results of measurements are often accompanied by confidence intervals. For instance, suppose a scale is known to yield the actual mass of an object plus a normally distributed random error with mean 0 and known standard deviation σ. If we weigh 100 objects of known mass on this scale and report the values ±σ, then we can expect to find that around 68% of the reported ranges include the actual mass.
If we wish to report values with a smaller standard error value, then we repeat the measurement n times and average the results. Then the 68.2% confidence interval is . For example, repeating the measurement 100 times reduces the confidence interval to 1/10 of the original width.
Note that when we report a 68.2% confidence interval (usually termed standard error) as v ± σ, this does not mean that the true mass has a 68.2% chance of being in the reported range. In fact, the true mass is either in the range or not. How can a value outside the range be said to have any chance of being in the range? Rather, our statement means that 68.2% of the ranges we report using ± σ are likely to include the true mass.
This is not just a quibble. Under the incorrect interpretation, each of the 100 measurements described above would be specifying a different range, and the true mass supposedly has a 68% chance of being in each and every range. Also, it supposedly has a 32% chance of being outside each and every range. If two of the ranges happen to be disjoint, the statements are obviously inconsistent. Say one range is 1 to 2, and the other is 2 to 3. Supposedly, the true mass has a 68% chance of being between 1 and 2, but only a 32% chance of being less than 2 or more than 3. The incorrect interpretation reads more into the statement than is meant.
On the other hand, under the correct interpretation, each and every statement we make is really true, because the statements are not about any specific range. We could report that one mass is 10.2 ± 0.1 grams, while really it is 10.6 grams, and not be lying. But if we report fewer than 1000 values and more than two of them are that far off, we will have some explaining to do.
It is also possible to estimate a confidence interval without knowing the standard deviation of the random error. This is done using the t distribution, or by using non-parametric resampling methods such as the bootstrap, which do not require that the error have a normal distribution.
One type of sample mean is the mean of an indicator variable, which takes on the value 1 for true and the value 0 for false. The mean of such a variable is equal to the proportion that have the variable equal to one (both in the population and in any sample). This is a useful property of indicator variables, especially for hypothesis testing. To apply the central limit theorem, one must use a large enough sample. A rough rule of thumb is that one should see at least 5 cases in which the indicator is 1 and at least 5 in which it is 0. Confidence intervals constructed using the above formulae may include negative numbers or numbers greater than 1, but proportions obviously cannot be negative or exceed 1. Additionally, sample proportions can only take on a finite number of values, so the central limit theorem and the normal distribution are not the best tools for building a confidence interval. See "Binomial proportion confidence interval" for better methods which are specific to this case.