6 Analyzing software-measurement data

6.1 Introduction

Data analysis involves several activities and assumptions:

• A data set or batch is a number of measurements from a number of entities.
• We expect the entities to be comparable so that we can examine the differences or similarities in the data.
• For this we wish to determine the characteristics of the attribute values for entities of the same type, or the relationships between attribute values for entities of the same or different types.
• To perform the analysis, we need statistical techniques.

6.2 Analyzing the results of experiments

In the book it is assumed that you understand basic statistics, including the following notions:

• measures of central tendency: mean, median, mode
• measures of dispersion: variance, standard deviation
• distribution of data: means are approximately normally distributed
• Student's t -test: a test to find out if two sample means differs significantly
• F-statistic: a test to find out if two sample variances differs significantly
• Kruskal-Wallis test: a test to find out differences between groups with rank data
• levels of significance: 0.05, 0.01
• confidence limits: interval around mean containing some % of the observations

We cannot in this course cover statistics properly, but will try to explain the most important things to take into account, and present some simple analysis techniques.