What is the difference between parametric and non-parametric stats tests?
Different kinds or levels of data need to be analysed using different kinds of stats tests. The basic distinction between using a parametric or non-parametric test is between simple data:
Nominal – this means “named” data such as male/female, adult/juvenile, or tallies/counts of data such as the number of students taking maths at a college.
Ordinal – this means that your data is ordered in some simple way such as sea lions being classified as small, medium and large;
and complex data:
Interval – this is the same as ordinal data except that we know, for example, the actual length of small, medium and large sea-lions.
Ratio – this is the same as interval data except that it must make sense to have zero of the variable being measured. For example it is quite possible to get zero out of 20 in a math test but you cannot have zero personality, or zero height.
You also need to consider the distribution of your data before deciding upon a stats test. To use a parametric test data need to be normally distributed (form a bell shaped curve when plotted) – you can get your stats software programme to plot data for you, and check that the “skew” and “kurtosis” equal 1. If the data are not normally distributed and the skew and kurtosis do not equal 1 then you need to use a non-parametric test.