chapter  15
28 Pages

More Complex Designs and The New Statistics in Practice

Researchers often use more than one measure and more than two conditions. Can the new statistics cope with realistically complex situations, such as factorial designs, multivariate analyses, and model fitting? That’s the first question I discuss in this chapter, and my answer is “increasingly, yes, although sometimes it takes extra effort.” My second question concerns the practicalities of adopting the new sta-

tistics, given all the pressures of the modern research world. I’ll discuss a number of strategies that may help. As a unifying theme I’ll use a general strategy for estimation that won’t

surprise you at all: Find the ES most appropriate for the research question, place a CI on it, then make a figure and interpret the ES and its CI. That’s the first section:

• A four-step general strategy for estimation

The reasons for preferring the new statistics to NHST are just as strong for complex designs as they are for the simple situations we’ve been discussing. I’ve chosen a few topics to illustrate how we can gain the advantages of the new statistics more widely. I’ll start with

• Analysis of variance (ANOVA) designs, especially randomized control trials (RCTs)

I’ll then take a side-step and discuss

• General issues that ANOVA designs raise, including multiple testing and data exploration

Then I’ll return to discuss further types of analysis:

• Model fitting • Multivariate analyses • Continuous data

Finally, I’ll move on to my second question and discuss strategies for putting the new statistics into practice.