In favour of randomized control trials
Torgerson and Torgerson (2001)
The dominant paradigm in educational research is based on qualitative methodologies (interpretive paradigm). Torgerson and Torgerson (2001. p. 317)
Despite fairly widespread use of quantitative methods the most rigorous of these, the randomised controlled trial (RCT), is rarely used in British educational research. Torgerson and Torgerson (2001. p. 317)
Even a well-designed quasi-experimental study is inferior to a well-designed randomised controlled trial. Torgerson and Torgerson (2001. p. 318)
Something I’ll need to get my head around – again!
The first problem is the statistical phenomenon of regression to the mean (Cook and Campbell, 1979; Torgerson, C.J., 2000) All groups studied need to have the same regression to the mean chances. Torgerson and Torgerson (2001. p. 318)
Non-randomised quantitative methods are nearly always inferior to the randomised trial. Torgerson and Torgerson (2001. p. 319)
Nearly 40 years ago Schwartz and Lellouch described two types of randomised trial: the ‘explanatory trial’ and the ‘pragmatic trial’ (Schwartz and Lellouch, 1967).
The explanatory trial design is probably the one with which most people are familiar.
This type of study is tightly controlled and, where possible, placebo interventions are used.
Thus, one may take a large group of children all from a similar socio-economic background and attainment and randomly allocate them into two groups. One group receives the intervention under investigation whilst the other receives a dummy or sham intervention. Torgerson and Torgerson (2001. p. 320)
The pragmatic trial : the environment in which the trial is conducted is kept as close to normal educational practice as possible.
The children, or schools, are allocated the new intervention at random. A disadvantage of the pragmatic approach is that the trials usually have to be much larger than the explanatory approach but the pragmatic trial approach is probably the most feasible and useful trial design for educational research. Because the trial mimics normal educational practice, there is a greater variation that can make it harder to detect a small effect. To cope with this the sample size needs to be increased accordingly. Torgerson and Torgerson (2001. p. 320)
The underlying idea of a randomised trial is exceedingly simple.
Two or more groups of children, identical in all respects, are assembled. Clearly, the individual children are different but when groups of children are assembled by randomisation, and with a large enough sample size, they will be sufficiently similar at the group level in order to make meaningful comparisons. In other words, the differences are spread equally across both groups, making them essentially the same. Torgerson and Torgerson (2001. p. 321)
The analysis of a randomised trial is actually simpler than other forms of quantitative research because we know the two groups are similar at baseline. Torgerson and Torgerson (2001. p. 321)
Randomisation creates groups with the same proportion of girls, with the same proportion of pupils from various socio-economic and ethnic groups, with the same distribution of ages, heights, weights etc. – this is the simple elegance of randomisation! Torgerson and Torgerson (2001. p. 323)
To avoid observer bias blinded outcome assessment must be undertaken. Torgerson and Torgerson (2001. p. 324)
Qualitative methodologies are well suited to investigating what happens with individuals; RCTs are appropriate for looking at the larger units relevant to policy makers. Torgerson and Torgerson (2001. p. 3264)
SCHWARTZ, D. and LELLOUCH, D. (1967) Explanatory and pragmatic attitudes
in therapeutic trials. Journal of Chronic Diseases 20, 637–648.
Torgerson, C, & Torgerson, D 2001, ‘The Need for Randomised Controlled Trials in Educational Research’, British Journal Of Educational Studies, 3, p. 316, JSTOR Arts & Sciences IV, EBSCOhost, viewed 13 February 201
- Effects of the Endpoint Adjudication Process on the Results of a Randomised Controlled Trial: The ADVANCE Trial (plosone.org)
- RCTs, skeptics, and evidence-based policy (errorstatistics.com)
- Smith and Pell: Parachute Use and Randomised Controlled Trials (delong.typepad.com)
Again, interesting references. The statement from Torgerson & Torgerson is reflected by innovation-guru Clayton Christensen. His work ‘Disrupting Class’ contains a chapter on educational research. See http://wp.me/p36s1C-4U for some reflections.