Statistics made simple--Part 3 statistical tests terminology.

Abstract

RSCN, MSc, CertEd(HE), DMS, Senior Lecturer in Children’s Nursing/Senior Nurse (Research Development), Glasgow Caledonian University/Yorkhill NHS Trust This is one of a series of short papers on aspects of research by Alison Twycross and Linda Shields You are reading some research papers and thinking about whether they might be relevant to your practice but they are full of statistical terms. What do these terms mean? Statistical terminology can be a bit confusing when you first encounter it. Within this paper we will endeavour to explain some basic statistical terms and identify what factors need considering when choosing which statistical test to use. The first question to answer is why do researchers use statistical tests? Inferential statistics are a way of establishing whether the results of a research study are due to chance effects or whether the results reflect real differences. Inferential statistics are used to answer questions such as: Does group A differ from group B? Is there a relationship between variable A and variable B? Inferential statistics are therefore used to test hypotheses (theories) about the relationships between two or more variables. Without carrying out statistical tests it is not possible to state categorically that a difference exists between two or more sets of data, or that a relationship exists between two variables. In quantitative research papers you will probably have noticed that researchers often state: ‘p 0.05) suggests that the result is more than 5 per cent likely to be due to chance. This result is said to be non significant, that is the results could be due to chance. However, a p value less than or equal to 0.05 (p< 0.05) suggests that there is a 5 per cent or less possibility of the results being due to chance thus indicating that there is a significant difference between the results, that is that the difference in the results is not due to chance. Statistical tests are therefore used to demonstrate whether results are statistically significant or not. But how do they decide which tests to carry out? One of the factors is the type of data that has been collected. Data can fall into one of three categories: nominal data, ordinal data, or ratio data. Nominal data refers to data which allocates responses into named categories (Hicks 1996). There are normally only two possible responses – for example, male or female, yes or no – which are non-numerical. This type of data is sometimes referred to as categorical data. Ordinal data allows the researcher to rank order participants’ responses along a dimension (for example, least pain – most pain). However, the differences between the points on the scale are not equal (Hicks 1996). Examples of ordinal data include Faces pain scales and the Glasgow coma scale. Ratio data are characterised by having equal intervals between the points of measurement. For example, when measuring a child’s blood pressure a difference of 10 means the same wherever it occurs – that is, a the difference between 80 and 90 is the same as the difference between 120 and 130. Examples of ratio data include blood pressure and temperature. This type of data is sometimes referred to as interval data. Another consideration when deciding which statistical test to use is whether the sample is independent or related. If an independent sample has been used data will have been collected from two distinct groups about the same variable – for example, heights of boys and girls aged eight years. However, if you have collected data from the same group of people on more than one occasion you have a related sample – for example, the height of ten eightyear-old boys each month for six months. If two or more separate groups have been closely matched to ensure that their characteristics which may affect the results are (virtually) identical this is treated as a related sample. PN