False-positive trial outcomes

Table 1.1 Expected trial performance (%) for various trial scenarios on the basis of precision and the probability of false- positive and falsenegative trial outcomes.

With all statistical tests, one aim is to ensure that, where there is no real effect, we will make false positive claims on only 5 per cent of occasions. Consider what would happen if there was actually no effect on clearance and we carried out 20 trials, each analysed by a one-sided test (testing for an increase). Bearing in mind that the actual procedure is to calculate a 90 per cent Cl, but then convert it to what is effectively a 95 per cent Cl by ignoring one of its limits, we can predict the likely outcomes as in Figure 10.2 ... 

In general, the evidence from animal studies suggests that more attention should be paid to the nature of the task the child is required to do in a test situation, and that as well as recording an outcome score in terms of correct responses, other measures of performance such as incorrect responses, false positives, speed of response, and trials to criterion should be recorded. For... 

---