Teaching strategies aiming to address:
Concern that statistical inference is hard to understand. Distributions of sample statistics underpinning calculation methods are
notoriously challenging concepts. Definitions are based on repeated random samples, not practical study designs. P-values are
often interpreted using language that does not accurately reflect the information they contain.
Concern over standards of statistical interpretation in the applied literature. Methods of statistical education and resulting varying
understanding of statistical inference contribute to this. Excessive focus on whether p < 0.05 leads to some people erroneously
declaring studies with very similar odds ratios as incompatible, merely because only one has p < 0.05.
Setting. Master’s in public health statistics course at Imperial College London.
Strategy. Refer regularly to calculated results amongst study participants and clarify that p-values and confidence intervals (CIs)
reflect imprecision resulting from random choices in selection of participants. Select CI and p-value “definitions” that clearly
reflect their purpose within them. Report the assumption of random sampling from some greater population throughout.
Include practical exercises that focus on graphs that further support conceptual understanding.
Conclusions. Statistics educators should steer away from common misconceptions by focussing on conceptual understanding of
CIs and of p-values. Research is warranted that assesses the subsequent impact of these methods on student understanding.