The different types of statistics tests and their main objectives 

There are many different tests in the field of statistics, here are the different types. 

The so-called "T-tests" will be used to compare data with each other. There are also tests designed to evaluate the potential relationships that may exist between two pieces of data. 

There are also quantitative and qualitative tests, parametric and non-parametric tests and finally tests that measure the relevance of data at a given time, this list is not exhaustive. 

In any case, these tests are used in multiple sectors of activity, companies can use them to analyze their marketing strategy, to help study customer satisfaction. 

Thus, statistical tests aim to compare data with each other, to find out if there are differences, when they occur and why. They also make it possible to understand how to associate two different data, to evaluate how often something happens, whether this frequency is stable, or "normal". Finally, statistical tests are also a way to prove research, theory or a hypothesis. Let us think of the academic field, but not only, since statistical tests can also be put forward in mathematics, medicine or any other field where it is possible or even necessary to formulate hypotheses. 

To choose a type of test, it is necessary to understand the field of application in which it will apply, as some are more specific than others. The more correctly the test is chosen, the more reliably the results can be analyzed. 

How to choose them? What are the main criteria for choice?

First, given the multiplicity of existing tests, it is important to understand what we are trying to analyze. If so, this could be a source of error for the conclusions that follow. 

What are the criteria sought? 

The first step in choosing a type of statistics test is to know the type of results and criteria you want to obtain.

The choice will be different depending on whether the data sought is qualitative or quantitative. In addition, it is also a question of determining whether the test in question will take the form of a satisfaction questionnaire, or whether it will be based more on physical characteristics

Choosing a statistical test can be more complex than it seems, given the choice but also because certain criteria can be grouped together when studying the data. 

 

How do you determine exactly whether the test you choose is going to be compatible with the data? 

It is a question of understanding whether the test in question relates to data that will be studied alone or in groups. 

The parametric test is often known as the t-test or ANOVA (analysis of variance). They are used, for example, to compare averages with each other.

The nonparametric test is for example the Wilcoxon, Mann Whitney or Spearman test. 

These two tests are not used in the same way. To explain things in a simple way, the first refers to homogeneous data while the second will apply more to more heterogeneous data, which fail to meet the conditions of normality. 

 

A different test depending on the type of data to be analyzed

To compare figures, averages: 

Student's t-test, or the one we mentioned above, ANOVA

Pearson and Fisher are statistical tests that are more regularly used to assess more qualitative data. 

To find out if a result conforms to a set of theories, tests such as Shapiro-Wilk or Kolmogorov-Smirnov tests are used more specifically.

 

Examples

Parametric Testing

Test t

A scientist wants to know if the students in group A are better than those in group B. So, he will establish the averages of the two groups and compare them to see if there is a real difference or not. 

Anova

In the same type of example, a researcher compares three types of diets. This type of test makes it possible to determine whether one of these methods allows for better results on the weight of the participants.

Test of Pearson 

For example, we are looking for a link between the number of hours of sports performed and the results obtained. The more hours of sport you have, the more the weight tends to drop, or not. 

 

Non-parametric tests 

Spearman

This type of test can be used to gauge customer satisfaction in a particular area, or for a certain brand. Customers are sometimes asked about the services known in the context of a restaurant service or in a department store, for example. On a scale of 1 to 5, they should rate a product or service. This type of test is increasingly being sent to consumers, especially online, as they only take a few seconds to respond, and they are very useful to the company in question. For example, evaluation of the reception of staff on a scale of 1 to 5, waiting time at the checkout, etc. The test is categorized as nonparametric because it does not follow a normal type of distribution. 

 

A final example for evaluating more qualitative data 

Chi carré

In a company, a manager wants to know if the success of a certain task depends on the age of the employees in a team. 

This type of test makes it possible to highlight or not a link between these two elements. 

 

Conclusion

The multitude of statistical tests available shows that it is very important to choose the right test in the right circumstances to avoid distorted results and conclusions. In various academic works, as in other sectors of activity, the choice of the right test can allow real advances in research and to obtain more relevant conclusions that are more in line with the initial theory. In the marketing sector, using the right elements and the right test can allow managers to improve the performance of companies or in the medical field for a doctor to improve the diagnoses made. 

 

https://pmc.ncbi.nlm.nih.gov/articles/PMC3116565/

https://www.scirp.org/journal/paperinformation?paperid=126875