Article Text

PDF

Making good use of statistics
  1. Darren Shaw

Abstract

Statistics are just a tool to help you answer the questions you have set. It may sound obvious, but the clearer the questions you set, the easier it is to answer them says Darren Shaw, who offers some advice on the use of statistics in practice research projects

Statistics from Altmetric.com

THE vast majority of my conversations with veterinary colleagues about the design and analysis of their research projects are about focusing in on what exactly the questions are.

Given all the variety out there, it is important to ask the right questions

Photograph: Cynoclub/Dreamstime.com

In the case of a project, the questions are normally framed in terms of the hypotheses you are testing. You need to ensure that your hypothesis is as clear and precise as possible. The more vague your hypothesis is at the start, the harder it will be to answer it in any meaningful way. For example, you need to be more precise than just, say, ‘Is there a difference in growth between male and female cats?’ You need to think about over what time period, what breeds, what part of the body is being measured, whether you need to control for diet/environment, etc. The more that you define exactly what you want to know – for example, ‘Is there a difference in growth as measured by the change in weight in the first two months of life between male and female non-pedigree kittens not weaned’ – the easier it is to set up the project, measure the correct things, analyse and present the results, interpret any results and, importantly, not to over-extrapolate. Darren Shaw is senior lecturer in veterinary clinical epidemiology at the Royal (Dick) School of Veterinary Studies

With respect to the actual statistics, a useful way to think about what statistics might be required is to use a framework to guide you. This will help you to decide how to summarise your data as well as from a statistical point of view. In addition, it will hopefully mean that if you do consult statistics books or other people, you have some idea of where you could be looking or what you should be asking. Otherwise you are faced with working through all the myriad different statistical procedures and tests out there trying to see which might be the one for the question you are trying to answer.

A number of possible frameworks can be used. The one that I use is to ask a series of questions about my data:

(1) How many groups will I have?

One (eg, one group of labrador retrievers); two (eg, male and female Siamese cats); more than two (eg, three different antibiotic treatment groups)?

(2) How many ‘things’ am I measuring?

One (eg, length of tail of puppies at six months old); two (eg, insulin levels and weight in 10-year-old cats)?

(3) How often have I measured these things? Once (eg, albumin levels at weaning); twice (eg, white blood cell counts pre- and three days post-treatment [ie, a paired sample where the same individuals are measured twice]); more than twice (eg, percentage of dermatitis in dogs measured once a fortnight for eight weeks)?

(4) Are my data:

Numerical (eg, tick burdens, length of the left forelimb) and within that are they (a) continuous/parametric (eg, creatinine levels) or (b) ranked/non-parametric (eg, body condition score, pathology index); or are my data categorical (eg, cat underwent surgery or chemotherapy) and within that are they (a) binary (eg, disease present/absent) or (b) more than two groups (eg, hair colour, breeds)?

Once I have answers to these four questions I can then select the appropriate statistical test based on the number of groups, and the frequency, type and number of different data I am measuring. Most statistical help pages/books will describe statistical tests in relation to these four questions.

Where to go for advice

A number of programmes are being run between some of the UK veterinary schools (Liverpool, Cambridge and Nottingham's Centre for Evidence-based Veterinary Medicine) and veterinary practices to encourage research in practice. In addition, a new initiative has been put together in consultation with BSAVA Petsavers where individuals at all of the UK veterinary schools have offered to help with queries about the design and proposed analysis of Petsavers projects prior to submission. E-mail inquiries can be sent to grantstudyadvice@bsava.com and queries will be forwarded to the appropriate vet school.

View Abstract

Request permissions

If you wish to reuse any or all of this article please use the link below which will take you to the Copyright Clearance Center’s RightsLink service. You will be able to get a quick price and instant permission to reuse the content in many different ways.