A minority of students did respond as though they applied the statistical test approach to inference.
In contrast, only a minority gave answers that were consistent with the statistical test approach.
In version A the statistical test approach was in favour of the disease and the likelihood ratio approach against, and in version B the pattern was opposite.
In both cases (Table 2), more respondents answered according to the likelihood ratio approach (A: 43.6%, B: 69.0%) than the statistical test approach (A: 7.5%, B: 4.5%).
The statistical test approach considers the probability of the observed result, and of more extreme results, under the null hypothesis - i.e., the p value.
Therefore the lack of endorsement by medical students of the statistical test approach to inference raises a red flag about this inference method.
Students who applied the likelihood ratio approach would interpret the test result as evidence for the disease, whereas those who applied the statistical test approach would fail to reject the null hypothesis, evidence against the disease.
We cross-tabulated the responses by type of scenario, comparing proportions of opinions that were compatible with the likelihood ratio approach, the statistical test approach, and the non-committed answers.
Thus if one applies the likelihood ratio approach, the test result argues against the presence of disease, and if one applies the statistical test approach, the test result leads to rejection of the null hypothesis of no disease, i.e., evidence for the presence of disease.
Within each cohort, three statistical testing approaches were considered.
