Data characteristics

Types of data characteristics

There are three types of data characteristics and they are as follows:

1. Non parametric (NP)
2. Non time series (NTS)
3. Time Series (TS)

• If data type is non time series (NTS) then you need to check it is normal or not, and if it is not normal then you need to convert it to normal to apply parametric test.
• For non-parametric (NP), if data is normal then apply parametric test and if data is not normal then you can apply nonparametric test.
• Time Series (TS) necessarily has to be normal. TS also have shocks, which will be studied later.

Tests for checking Normal Distribution:

To check the normal distribution of our data / observation, there are two test:

1. For parametric data SHAPIRO test can be applied.
2. For non-parametric data Kolmogorov–Smirnov test or K Smirnov test.

Why to check normal distribution in NT-NP Series?

Ans: To check which econometric tool to apply. Although, NT-NP does not follow any distribution, but still we have to check normal distribution and if it is then we will apply parametric test else non-parametric test will be applied.

Why to check normal distribution in NT – Parametric series?

Ans: To decide whether the data has to be transformed or not.

Why to make NT – parametric series normally distributed?

Ans: The answer to this question can be understood through the study of OLS (Ordinary Least Square) model and Blue model, which will be our next topic of discussion.

 OLS (Ordinary Least Square) Vs Blue Model:

The above model shows the mathematical representation of OLS Model. OLS model is a method in which the independent variable predicts the dependent variable in a linear pattern.

In the above expression, the dependent variable (DV) should be normal so that the expression can be unbiased. If the dependent variable is not normal or it is biased, then we need to transform it to make it unbiased and normal. After transformation this OLS model will then become Blue. Therefore, OLS model will be blue when dependent variable will be unbiased and hence will be unbiased only when dependent variable will be normally distributed. This answers to the question: “Why to make NT – parametric series normally distributed?”