STATISTICS


The correct starting point is: Are we comparing one group with a fixed target, two groups, or more than two groups? Are the observations independent or paired? Is the outcome continuous, ordinal, binary, or categorical? Are we interested in a difference in means, medians, ranks, or distributions? Are there repeated measurements on the same samples, patients, staff members, days, or instruments? Is the observed difference large enough to mater in laboratory practice? Was the comparison planned before seeing the data, or selected afterwards?


Failure to answer these questions can lead to technically neat but scientifically weak analysis.


What is hypothesis testing trying to answer? Before considering individual tests, it is worth stepping back and asking what hypothesis testing is for. It is a formal way of asking whether the observed data are compatible with a particular statistical model. In many routine examples, that model is the ‘there is no difference’ model. For example: no difference in mean turnaround time before and after a workflow change no difference in mean result between fresh and stored specimens no difference in mean competence score before and after training no difference in mean result across three storage temperatures no difference between the observed mean and a predefined target.


The usual starting point is the null hypothesis. This is often described as the hypothesis of no difference, no change, or no effect. The alternative hypothesis is that there is a difference, change, or effect. The test then compares the observed difference with the amount of variation expected if the null hypothesis were true and the assumptions of the test were reasonable. The result is often summarised as a


P-value. A P-value is not the probability that the null hypothesis is true. It is also not the probability that the result occurred by chance. More accurately, it


Test One-sample t-test Paired t-test


Unpaired Student t-test/ independent t-test


Welch’s t-test Table 2. t-test variations.


indicates how incompatible the observed data are with the statistical model being tested, assuming that the model and its assumptions are correct. A conventional threshold such as


P<0.05 is therefore not a scientific truth boundary. It may be useful, but it should not replace judgement. A result just below 0.05 is not automatically important, and a result just above 0.05 is not automatically unimportant. Those decisions require effect sizes, confidence intervals, predefined criteria, knowledge of the measurement procedure, clinical context, and professional judgement. It is also important to recognise when


hypothesis testing is not the main question. If the real question is whether or not a process meets an audit standard, the proportion of cases meeting the standard may be more relevant than a t-test. If the question is whether or not two conditions are sufficiently similar, absence of statistical significance is not enough; equivalence or non-inferiority thinking may be required. If the question is whether or not a measured difference is acceptable, the confidence interval should be interpreted against predefined limits rather than judged only by P<0.05.


What are some commonly used hypothesis tests?


The t-test(s)


A t-test compares a mean, or a difference between means, accounting for the variation expected from random sampling. It is used when the outcome is continuous and the comparison is focused on means.


Poor test selection can lead to false reassurance, unnecessary process changes, or overconfident conclusions from limited data


The origins of the t-test are a useful


reminder of its practical purpose. William Sealy Gosset developed the work behind what became known as Student’s t-test while working at Guinness, where small-sample experiments were needed to support industrial quality decisions. He published under the name ‘Student’, and the method became associated with making cautious inferences from limited data. That history is directly relevant to laboratory medicine: many laboratory audits, verification exercises, training evaluations, and service studies also involve modest sample sizes and practical decisions made under uncertainty. The most common versions are shown in Table 2.


The one-sample t-test The one-sample t-test is sometimes overlooked, but it is conceptually important. It is used when there is one observed group and the comparison is with a fixed value. That value might be a target, claimed value, historical value, expected value, or theoretical value. Laboratory examples include things like testing whether a mean difference differs from zero, or a mean recovery differs from 100%.


The key requirement is that the


comparator value (ie zero or 100% in the example above) must be defined independently of the data being analysed. It should not be selected after looking at the results. Showing that a mean is statistically


different from a target is not the same as showing that the result is unacceptable. For example, if the mean recovery is 98.5% and the target value is 100%, a large dataset may produce a statistically significant difference. That does not automatically mean the recovery is unacceptable. If the predefined acceptable range is 95–105%, the observed recovery may be entirely acceptable despite being statistically different from 100%. The opposite problem also occurs.


June 2026 WWW.PATHOLOGYINPRACTICE.COM 17 Used for


Comparing one sample mean with a fixed target or hypothesised value


Comparing two related measurements


Comparing two independent


group means when equal variance is a reasonable assumption


Laboratory example


Is the mean recovery different from 100%?


Same specimens tested before and after storage


Turnaround time from two independent request locations with similar variability


Comparing two independent means Two groups with unequal SDs without assuming equal variance


or unequal sample sizes


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