Professional Documents
Culture Documents
probability distribution early in their courses. One of thebest known is the normal
and Gaussian family, where individual members are specified by assigning constant
values to two quantities called parameters. These are usually denoted by and
and represent the mean and variance. The notation N (, ) denotes
a normal distribution with these parameters. Normal distributions are often
associated with continuous data such as measurements. Given a set of independent
observations (a concept explained more fully in section 1.2) froma normal
distribution, we often want to infer something about the unknown parameters. The
sample mean provides a point (i.e. single value) estimate of the parameter
(sometimes called, in statistical jargon, the population mean). Here the wellknown t-test is used tomeasure the strength of the evidence provided
by a sample to support an a priorihypothesized value for the population
mean. More usefully, we may determine what is called a confidence interval for the
true population mean. This is an interval for which wehave, in a sense we describe
in section 1.4.1, reasonable confidence that it contains the truebut unknown mean
. These are examples of parametric inferences. The normal distribution is strictly
relevant only to some types of continuous scale data suchas measurements, but it
often works quite well if the measurement scale is coarse (e.g. forexamination
marks recorded to the nearest integer). More importantly, it is useful
forapproximations in a wide range of circumstances when applied to other types of
data. The binomial distribution is relevant to some types of counts. The family
also has twoparameters n, p, where n is the total number of observations and p is
the probability that aparticular one from two possible events occurs at
any observation. Subject to certainconditions, the number of occurrences, r,
where 0 < r < n, of that event in the n observationshas a binomial distribution,
which we refer to as B (n, p) distribution.