# Non Parametric

Parametric statistics uses certain assumptions and key information for analysing the data. The formulas have been created on the assumption that the distribution is normal or that the subjects have been suitably ranked. However, there are instances where this information is not available, and the data set is not normal. In such cases, one has to use Non Parametric statistics. Non Parametric statistics makes use of formulas which takes into account the fact that the data does not have any clear structure or definable characteristics.

Let us say that you conducted a survey. You asked the respondents to give a ranking based on 1 to 7. When you have collected this data, you would be able to use parametric methods for carrying out the analysis. In case you had merely asked the respondents to give yes and no answers then you would have to use non parametric methods. Parametric methods need more information and assumptions to be successful than non parametric methods. But this also means that the inferences made by non parametric methods would be applicable in a greater number of scenarios since they are not constrained by those parameters.

When you are looking through a data set, it is essential that you understand what type of statistics would give a reliable answer. If the measurement scale is ordinal, then it is best to use non parametric statistics. On the other hand if ratio scales are being used then the help of parametric statistics should be taken. However, most statistical tests have both parametric and non parametric analogs.

One of the problems with non parametric tests is that you tend to lose power over the analysis. This is because the data does not have the fineness that is needed to perform an in-depth analysis. If you asked people to rate their likes and dislikes on a scale of 1 to 5, then you would be able to measure the fineness in their responses. On the other hand if you simply asked them to rate like or dislike, then the responses would become generalised. Considering that the whole purpose of statistics is to find minute patterns, non parametric analysis puts the statistician at a bit of a loss. These are factors to consider when designing a survey, for instance. Some of the commonly used non parametric methods are the Mann Whitney test, Kruskal Wallis test, Sign test, Wilcoxon test and the Friedman test.