ABSTRACT
The teacher decided to teach half the class on the New Scheme and half on the Old Scheme on the next class intake. The children were randomly allocated to the two schemes, to avoid biasing the samples due to factors such as intelligence. In this way the two samples were assumed to systematically differ only on the variable under study: reading scheme. The teacher can now compare the samples. Yet the teacher is not really interested in the samples as such but the population of children these samples are drawn from. Is the New Scheme better for children of this age rather than just this class? The question is whether the population distribution for the New Scheme is higher up a scale of reading performance than the distribution for the Old Scheme. This is a one-tailed prediction that the New Scheme will result in better performance than the Old Scheme. Unfortunately the teacher has no details of these populations, they are both unknown.
