Unbalanced and confounded designs
This chapter deals with techniques for removing differences between the pre-existing groups by adjustment. The researcher may have to allocate the different conditions to pre-existing groups defined by, for example, attending the same health centre, being pupils of the same school class or working on the same production line. Research can lead to conclusions about one variable causing an effect on another variable. Equal numbers of individual subjects in each condition or combination of conditions will not always be possible in real-world research. The same problem of unequal numbers of subjects in groups is an inevitable part of research in which there is no random allocation or intervention by the researcher. The Home Office has commissioned research on the effect of type of conviction on locus of control scores on prisoners. The locus of control scores is low for ‘external’ and high for ‘internal’ individuals.