The methods of classical statistics evaluate the possibility that when a result looks interesting, it is merely the uninteresting variability in the measurements that has caused it. Individual values departing considerably from the rest of the data can have a marked effect on some statistical calculations. Such exceptional scores sometimes have different origins from the others with which they have been grouped and may be considered to be outliers, that is, scores that do not properly belong with the others. The descriptive approach called parameter estimation focuses on the size of effect rather than on significance. Parameters are hypothetical characteristics of the population and they are estimated from the data. If the data are consistent and plentiful, the interval will be narrow so any difference between the treatments must be small. The reason is that with a large amount of data, significance can be demonstrated even for a very weak effect.