How to Start Analyzing, Test Assumptions and Deal with that Pesky p-Value
This chapter discusses the steps to take before any of the analyses discussed in earlier chapters. Although it may seem counterintuitive to put this information in the last chapter, experience teaches us that these are things people do not want to read fi
- PDF / 402,725 Bytes
- 22 Pages / 439.37 x 666.142 pts Page_size
- 58 Downloads / 207 Views
How to Start Analyzing, Test Assumptions and Deal with that Pesky p-Value
This chapter discusses the steps to take before any of the analyses discussed in earlier chapters. Although it may seem counterintuitive to put this information at the end of this book, experience teaches us that these are things people do not want to read first when they embark on their analysis journey. We all start out with a big idea and full of courage, but all too often our courage is blown to bits because words and terms like “homoscedasticity,” “skewness,” and “multivariate normality” make our heads spin and our plans seem impossible. However, we hope that, after you have gotten a kick from seeing first results with the method of your choice, you are now ready to learn about all the things you should have done first—the things that make your results credible. No data is perfect, but understanding how imperfect your data is, correcting the most important imperfections, or using a different method will help you to obtain credible results from your imperfect data. The first step on the journey is to structure your data in a way that best fits your research questions and analysis plan. Then some thorough cleaning will rid your data of imperfections in the details before you start to understand the larger imperfections. These larger imperfections and the relationships in your data are first explored by summarizing and visualizing data. These are the topics of Sect. 7.1. After that, we have a more thorough look at the possible larger imperfections by discussing how to test the important assumptions with which your data must align. The four groups of assumptions we discuss are independence, normality, homogeneity of variance, and linearity. We discuss how to test whether these assumptions hold for your data and briefly introduce strategies for when they do not. Finally, we share some of the latest thinking and our perspective on how to deal with that pesky p-value. This issue of the correct use of the p-statistics is prevalent and ubiquitous, and you are well served to follow the debate and stay current in this regard.
# Springer International Publishing Switzerland 2017 W. Mertens et al., Quantitative Data Analysis, DOI 10.1007/978-3-319-42700-3_8
135
136
8.1
8
How to Start Analyzing, Test Assumptions and Deal with that Pesky p-Value
Structuring, Cleaning, and Summarizing Data
As we mentioned in the introduction to this book, analyzing data typically starts with a phase of structuring, cleaning, and exploring. You could call this process “sensemaking” of the data—working to understand what you have. Extracting meaning from data requires it to be structured in a logical and consistent way and ridding it of unreliable and invalid data. Once that is done, we can start exploring and summarizing the data in statistics and graphs. Let’s discuss each of these steps in turn.
8.1.1
Structuring Data
Step 1: Structure Cases in Accordance with Your Research Question(s) The first step is to structure data in a format that allows you to run
Data Loading...
