The trials of evidence-based education: the promises, opportunities and problems of trials in education

Click to increase image sizeClick to decrease image size

Notes on contributor

Matthew Inglis is a Reader in Mathematical Cognition in the Mathematics Education Centre at Loughborough University. He is interested in understanding the processes involved in numerical and mathematical thinking, and how these can be promoted through education.

Notes

1 For a balanced account of the three main approaches to statistical inference, I recommend Dienes’s (2008) excellent book.

2 Gorard et al. write: “a power calculation starts by envisaging a non-zero effect size (the estimated effect of the treatment in an RCT). The researcher assumes this non-zero effect size as the bedrock for the calculations that follow. The calculations themselves are also predicated on a significant test, which was shown in Chapter 3 to assume as the basis for its own calculation that there is no effect sizes (the nil-null hypothesis). Put another way, the p-values generated by significance tests assume an ES of precisely zero. Both of these initial assumptions cannot be true in the same calculation, by definition. Therefore, ‘power’ does not make sense” (p. 40).