Plain Language Translation of Statistical Concepts
Statistical communication fails most often not because the statistics are wrong, but because the language used assumes expertise the audience does not have — and sometimes because experts themselves use language imprecisely. Key translation strategies: p-values: 'The results were statistically significant' is meaningless to most audiences. Better: 'We found strong evidence that the treatment works — results this large would only happen by chance about 3 times in 100 if the treatment had no effect.' Even better: focus on the effect, not the p-value — 'People who took the treatment lost an average of 4.2 kg more than the control group over 6 months.' Effect sizes: replace r = 0.43 with 'moderate positive relationship' or describe variance explained: 'Exercise habits account for about 18% of the variation in depression scores.' Confidence intervals: 'We are 95% confident that the true average difference is between 2.1 and 6.3 units' — avoid 'the true value is definitely in this range.' Risk and probability: absolute risk is more informative than relative risk. 'The drug reduces heart attack risk from 2% to 1%' is more informative than 'the drug cuts heart attack risk in half.' The latter (relative risk reduction of 50%) sounds dramatic; the former (absolute risk reduction of 1%) is more honest. Number Needed to Treat (NNT) = 1 / ARR: NNT = 100 means you must treat 100 patients to prevent one heart attack — provides intuitive clinical context.