Statistics made simple. Part 4. Choosing the right statistical test.

Correction In the July/August edition of Paediatric Nursing (16, 6, 36) author details for Linda Shields should have been included This is one of a series of short papers on aspects of research by Alison Twycross and Linda Shields A previous paper in this series explained some statistical terms and why carrying out statistical tests is important. It also highlighted some of the issues to consider when choosing which statistical test to use (Twycross and Shields 2004). Within this paper the appropriate use of different statistical tests is described. Before deciding which statistical tests should be used with a set of data you need to consider: ❘❚❘ whether the data collected were nominal, ordinal or ratio ❘❚❘ the number of groups used in the study ❘❚❘ whether your sample is independent or related ❘❚❘ whether you are trying to establish a difference between group A and group B ❘❚❘ or whether you are trying to ascertain the relationship (correlation) between two or more sets of data. If you are looking at differences between groups you need to use a test for statistical differences which gives you a p value. However, if you are interested in whether there is a relationship between two variables (described in statistical terms as a correlation) you need to use a statistical test that tests for correlation. This will give you a correlation co-efficient (r). Correlations can be positive or negative. A positive correlation is when high scores for one variable are associated with high scores in the other, whereas a negative correlation is when high scores for one variable are associated with low scores in the other (Hicks 1996). In Table 1 the tests used to ascertain whether there is a statistical difference between groups of data are in black. Those used to test for correlation are in blue. Table 1 identifies which type of data (nominal, ordinal or ratio) and sample (independent or related) are appropriate for each test. If you carry out a test to ascertain whether there is a statistically significant difference between two or more groups of data you need to examine the p value obtained. A p value less than or equal to 0.05 (p ≤0.05) indicates a significant difference. If you are trying to establish whether there is relationship (correlation) between two variables you need to look at the correlation co-effecient (r) and the p value. The p value will tell you whether the correlation is statistically significant whereas the correlation co-efficient (r) tells you the strength of the relationship between the two variables. A correlation coefficient will be between -1 and +1 (a correlation of 0 means no correlation). The nearer to 1 the value is, the stronger the relationship. Thus if r = 0.9 there is a strong positive correlation between the two variables, whereas if r = -0.2 there is a weak negative correlation PN