STATISTICS


Common mistake Starting with the test


Treating hypothesis testing as the whole analysis


Confusing difference from a target with acceptability


Ignoring paired data


Using repeated t-tests for multiple groups


Treating non-parametric tests as assumption-free


Reporting only P-values


Interpreting ‘not significant’ as ‘no difference’


Forgeting practical significance Using normality tests mechanically Leting software defaults decide


Choosing a one-tailed test after seeing the result


Why it maters


The analysis may not answer the laboratory question.


A P-value does not replace plots, summaries, effect sizes, confidence intervals, or judgement.


‘Significantly different’ does not always mean


unacceptable; ‘not significant’ does not prove acceptable. Treating linked results as independent can give


misleading conclusions.


Multiple testing increases the risk of false-positive findings.


Rank-based tests still have assumptions and may answer a different question.


Statistical significance does not show the size or importance of an effect.


A small or noisy study may simply lack power.


A statistically significant result may be too small to mater.


Normality testing alone should not dictate the method.


Software cannot know the laboratory question or data structure.


This can make evidence appear stronger than it is. Table 3. Some possible pitfalls if correct choice of statistical test is not undertaken.


produce higher values than another, but it may not directly estimate the mean concentration difference that maters for a specific analytical claim. Similarly, a statistically significant ANOVA result may show evidence of difference somewhere among the groups, but it does not identify the cause or determine whether the difference is operationally acceptable (Table 3).


Calculators


Here are some calculators to help with some of your analytical work.


htps://


pathologyuncertainty.com/ calculators/


Further reading MacDonald S. Distribution, outliers and


confidence intervals in laboratory data. Pathology in Practice 2026 May; 27 (3): 22–6.


Dr Stephen MacDonald is Consultant Clinical Scientist, The Specialist Haemostasis Unit, Cambridge University Hospitals NHS Foundation Trust, Cambridge Biomedical Campus, Hills Road, Cambridge CB2 0QQ.


+44 (0)1223 216746 20 WWW.PATHOLOGYINPRACTICE.COM June 2026


Conclusions It is clear that software-driven statistics can be very helpful (when looking at the maths involved if we had to do it by hand! – Fig 1), but come at a cost that can quickly outdo the benefit of point and click interfaces. Just because we can run a statistical test, doesn’t mean we should. Statistical assumptions, study design and data quality are central to us making


the correct decision for what test to run, if at all! This is equally important for operational data in the laboratory, as it is in the research seting. If you are doing a research project, don’t collect your data then ask your colleagues what test you should be running – by that time it is too late.


It’s not to say that software is a bad thing – of course it isn’t, and in an ironic twist I have again developed a learning tool for one sample t-test that shows the complexity of the underlying assumptions with what is a simple test to run normally. If you are bored, feel free to have a play about with the app and see if it helps – there is lots of information about the mechanics of the calculations available in there. It is available at: htps:// pathologyuncertainty.com/calculators/ PPi


Beter approach


Define the question, comparison, outcome, and design first.


Use the test as one part of interpretation.


Compare results with predefined acceptability limits and uncertainty.


Use paired analysis for split samples, before/after data, or repeated measurements.


Use ANOVA, Welch ANOVA, Kruskal–Wallis, or planned comparisons as appropriate.


Choose them only when their interpretation matches the laboratory question.


Report the observed difference and confidence interval.


Treat non-significance as inconclusive unless equivalence was specifically tested.


Interpret findings against clinical, operational, or predefined criteria.


Use plots, study design, sample size, outliers, and assumptions together.


Select the test deliberately based on design and interpretation.


Use one-tailed tests only when justified and specified in advance.


Page 1  |  Page 2  |  Page 3  |  Page 4  |  Page 5  |  Page 6  |  Page 7  |  Page 8  |  Page 9  |  Page 10  |  Page 11  |  Page 12  |  Page 13  |  Page 14  |  Page 15  |  Page 16  |  Page 17  |  Page 18  |  Page 19  |  Page 20  |  Page 21  |  Page 22  |  Page 23  |  Page 24  |  Page 25  |  Page 26  |  Page 27  |  Page 28  |  Page 29  |  Page 30  |  Page 31  |  Page 32  |  Page 33  |  Page 34  |  Page 35  |  Page 36  |  Page 37  |  Page 38  |  Page 39  |  Page 40  |  Page 41  |  Page 42  |  Page 43  |  Page 44  |  Page 45  |  Page 46  |  Page 47  |  Page 48  |  Page 49  |  Page 50  |  Page 51  |  Page 52  |  Page 53  |  Page 54  |  Page 55  |  Page 56