Statistical analysis

You have collected data on the blood pressure of 40 patients both pre- and postadmission

to the critical care unit. You now wish to analyse this data. Which of the

following statistical tests would be the MOST APPROPRIATE?

a) Mann-Whitney U test

b) Spearman’s rank correlation coefficient

c) Kruskal–Wallace one-way analysis of variance

d) Paired Student’s t-test

e) Wilcoxon signed-rank test

Answer: d

Explanation

There are many examples in medicine of continuous data (blood pressure, cholesterol,

ejection fraction, etc.) There are a number of statistical methods for analysing this data,

which can be divided into two main groups based on whether or not assumptions are

made about the data that is being examined. Some distributions of data are described

by quantities called parameters, for example the mean, standard deviation or variance,

and methods that use distributional assumptions are called parametric methods. These

methods (including various t-tests, and analysis of variance for comparing groups –

also known as ANOVA) make the assumption that the data are normally distributed

and also that the spread of the data is uniform either between the groups or across the

range of values being examined. Non-parametric statistical tests (including those in

Options (a), (b), (c) and (e)) do not require the data to follow a particular distribution

and work by analysing the rank order of the data rather than the measurements

themselves. Broadly speaking they are less powerful statistical tests unless the number

of observations involved is large. It is possible to convert non-normally distributed

data to make it suitable to be analysed by the more powerful parametric statistical tests.