When conducting research, it is important to reach accurate conclusions and generalizations concerning the participant groups. For this purpose, statistical tests are used, which can be classified into parametric and non-parametric. While for the first type, the researcher is aware of the parameter to be applied to the sample, the second type implies that no specific parameter has been identified (Hannigan & Lynch, 2013).
A parametric test is a type of statistical hypothesis test that allows making generalizations and inferences about the parameters (or defining properties) of the population from which the researcher draws data. Examples include one- and two-sample tests (t-tests, z-tests, the analysis of variance, paired tests, etc.). A t-test is one of the most commonly used parametric tests based on t-statistics; in running this test, the researcher should rely upon the assumption that the identified variable has a normal distribution. The interval scale is applied to measure the variables of interest, and the population variance is to be calculated for the given sample. Parametric tests are used when more statistical power is required, or the spread of each group is different (Hannigan & Lynch, 2013).
In contrast, a non-parametric test can be defined as a statistical hypothesis test that does not rely on particular parameters set for the population distribution. This type of test is often referred to as a distribution-free test based on differences in medians. All variables are measures on a nominal level (applied for non-metric variables). Examples include chi-square tests, the Fisher Exact Probability test, the Mann-Whitney Test, the Friedman Test, and more. These are applied when the area of study is better represented by the median, when the sample size is small, and when ordinal and ranked data is involved (Collett, 2015).
The major differences between the underlying assumptions of the two tests can be summarized as follows (Collett, 2015):
- A statistical test that relies on several assumptions concerning the parameters is called a parametric test, whereas one used for non-metric independent variables is referred to as non-parametric.
- In the case of a parametric test, distribution is the major basis for statistics, while a non-parametric test uses arbitrary statistics.
- In a parametric test, the measurement is performed on a ratio or interval level; in contrast, in a non-parametric test, the ordinal scale is used.
- The measure of central tendency is the mean in a parametric test and the median in a non-parametric test.
- The information about the population is complete in a parametric test, whereas in a non-parametric test, no information is provided.
- A parametric test can be applied only for variables, while a non-parametric test is also suitable for attributes.
- Pearson’s coefficient of correlation is used in a parametric test, and Spearman’s rank correlation is used in a non-parametric situation.
References
Collett, D. (2015). Modelling survival data in medical research. Boca Raton, FL: CRC press. Web.
Hannigan, A., & Lynch, C. D. (2013). Statistical methodology in oral and dental research: Pitfalls and recommendations. Journal of Dentistry, 41(5), 385-392. Web.