Statistical approaches in surface finishing. Part 1. Introductory review and parametric hypothesis testing

Abstract

This paper is the first of a short series of articles aimed towards describing some of the various statistical methods and approaches that have been used in surface finishing. The methods fall broadly into two areas: analysis and design-of-experiments. This article introduces the subject, briefly reviewing the wide use of a number of experimental design tools in recent surface finishing research before starting with a discussion of parametric hypothesis testing, the simplest of the statistical methods.

Statistical approaches in surface finishing. Part 2. non-parametric methods for data analysis

“…The Mann-Whitney U test is the non-parametric equivalent of the Student's t-test, 5 which cannot be applied due to the data not being normally distributed. * Results of the antimicrobial and antifungal behaviour of interior coatings with nano ZnO were confirmed repeatedly, hence the Wilcoxon rank-sum test, which assesses whether two independent samples of observations come from the same distribution.…”

“…There has been a marked increase in the application of sophisticated statistical tools in both the design of materials finishing experiments and in the analysis of data in recent years. [1][2][3][4] Examples from the first category usually come under the design-of-experiments (DOE) type, which were briefly summarised in Part 1 of this series; 5 fundamental and methodological details describing the use of DOE will follow later in this series. Part 1 also contained a 'beginner's guide' in how to perform key parametric statistics tests using the stats software package SPSS.…”

“…Part 1 also contained a 'beginner's guide' in how to perform key parametric statistics tests using the stats software package SPSS. 5 Parametric tests assume data are independent, normally distributed and have homogeneity of variance.…”

“…Many of these have parametric equivalents, some of which were discussed in Part 1. 5 Before going into detail on how to perform each of the chosen non-parametric tests, some selected examples from the literature where such tests have been applied to analyse real data from surface finishing/coatings experiments are considered ( Table 2). The reader is reminded that the dependent variable (dv) is the measured property where the effect caused by the intervention is expected, for example, surface roughness, thickness, leach rate, hardness, % elongation at break, reflectance etc; this property is usually plotted on the y-axis of a graphical display.…”

“…The independent variable (iv) is normally the group type, usually found on the x-axis, examples being coating type, alloy composition, cell type, exposure environment etc; this property can also be continuous data, e.g., time. The type of non-parametric test to be employed depends on the type of data (interval, ordinal or nominal), 5 and the number of groups (Table 1). The various tests are described and, for some, example analyses are given to assist the reader in understanding fully the application.…”