1: \begin{abstract}

2:   A key aspect of the design of evolutionary and

3:   swarm intelligence algorithms is studying their

4:   performance. Statistical comparisons are also a crucial part which

5:   allows for reliable conclusions to be drawn. In the present paper we

6:   gather and examine the approaches taken from different perspectives

7:   to summarise the assumptions made by these statistical tests, the

8:   conclusions reached and the steps followed to perform them

9:   correctly. In this paper, we conduct a survey on the current trends

10:   of the proposals of statistical analyses for the comparison of

11:   algorithms of computational intelligence and include a description

12:   of the statistical background of these tests. We illustrate the use

13:   of the most common tests in the context of the Competition on

14:   single-objective real parameter optimisation of the IEEE Congress on

15:   Evolutionary Computation (CEC) 2017 and describe the main advantages

16:   and drawbacks of the use of each kind of test and put forward some

17:   recommendations concerning their use.

18: 

19: %   We examine the application of some non-parametric tests and advanced

20: %   proposals, like the use of Page test to study the convergence of the

21: %   algorithms. The main contribution of this paper is an extensive

22: %   guideline on the use of the statistical tests along with the

23: %   description of the use of some existent Bayesian test presented in

24: %   classification algorithms comparison context to optimisation

25: %   algorithms comparison scenario, with the advantages and the qualms

26: %   of each test. 

27: \end{abstract}

28: