Statistics are numbers which present information about data. They show patterns in the data which may represent tendencies and trends within the research population. They show the impact of different variables on a population.

Statisticians use a range of tests to test the significance of numerical results. They are actually testing whether the differences between sets of figures are real differences or not. The type of test will differ depending on the type of numerical data, ie, whether it is from a nameable or measurable characteristic, whether it is nominal or ordinal and if the latter if it is measured according to an interval scale or a ratio scale.

• The Chi square tests the independence or interdependence of nameable characteristics (variables) within a population. So if you have variable x and variable y you can find out what proportion of the population use both x and y.

• The t-test is used to test whether two samples are from the one population or not, that is do they do the same thing or are they significantly different and therefore constitute a separate sample group. The characteristic used in the t-test must be measurable and the scale type interval or ratio.

• Analysis of variance (ANOVA) is of two types: the one-way design and factorial design.

  • With the one-way design the role is the same as the t-test, to test whether samples are from the same population or represent different populations with respect to some measurable characteristic, but unlike the t-test which only tests two-samples at a time, the one-way design ANOVA can test any number of samples provided they represent different levels of the same characteristic.
  • Factorial design ANOVA uses the means of two or more nameable characteristics to ask if there is more variation between groups scores than within group scores.

• A correlation measures the extent to which characteristics or variables vary together within a population. There are two types of correlation test:

  • Pearson product moment correlation measures characteristics of the one sample and both characteristics must be measurable on an interval or ratio scale.

  • Spearman rank order correlation measures two characteristics of the one population in an ordinal scale.