Statistics Guide for Research Grant Applicants

F. General

F-1 Statistical expertise (statistical analysis)

To some it may seem pointless to describe in detail a proposed statistical analysis for data that are not yet in existence. It might be argued, that at this stage the correct approach is difficult to judge, as without the data you cannot be sure of distributional properties (e.g. whether a variable follows a Normal distribution) and therefore assumptions. However, all too often statisticians are presented with data that they cannot analyse. The flaws in design which lead to this problem only come to light when someone starts to think about how the data can be analysed. How much better for all concerned if these flaws come to light at the proposal stage. If applicants are not sure how to analyse their data, they should consult a statistician.

Thinking about the statistical analysis or discussing it with a statistician brings home to some researchers the need to buy in statistical expertise. How much statistical input is required may vary from a few days of a statisticians time to enlisting a statistical collaborator /co-applicant. Where the statistical methods are complex (e.g. multilevel modelling) the statistical reviewer will be looking for some reassurance that the research team has the ability to complete such an analysis and to interpret the results appropriately; clearly a statistical collaborator would provide that reassurance.

Back to top

F-2 Statistical software

Applicants need to make sure that they have access to the right statistical software or that a request for the correct software is included in the grant proposal. There are many commercial statistical software packages, e.g. STATA, SAS, SPSS, GENSTAT, StatXact, MLWin, S-PLUS, BMDP, MINITAB, CIA, etc., but they do not all do the same things and they vary both in their flexibility and in their complexity of operation. Free statistical software is also available in the form of Epi Info, which can be downloaded from the web.

Back to top

F-3 Ethics

F-3.1 The misuse of statistics

Altman (1991, 1982) argued forcefully that the misuse of statistics was unethical. The term misuse of statistics covers both poor study design and inappropriate statistical analysis. These are unethical as they can lead to conclusions that are erroneous or misleading. At best, patients are inconvenienced and resources squandered for no good reason. At worst patients are harmed through inappropriate clinical decisions based on the erroneous research results.

Back to top

F-3.2 Critical appraisal

Altman (1991, 1982) extends his ethical debate stressing the need for the results of previous studies to be reviewed critically and not taken on face value. Without critical appraisal erroneous results may lead to further research in a totally unhelpful direction. This observation is particularly pertinent to grant applicants when compiling the Background section of any research proposal. Results of previous research should not be presented uncritically.

Back to top

F-3.3 Studies involving human subjects

Studies involving human subjects, especially clinical trials, raise many ethical issues. Guidance on these issues is contained in the Declaration of Helsinki. This declaration of ethical principles was first adopted by the World Medical Association at its General Assembly Meeting in Helsinki in 1964. It was updated in 1975, 1983, 1989, 1996, 2000, 2004, and most recently in 2008. It is essential reading for any medical researcher, especially those planning to conduct a clinical trial. Indeed mainstream journals consider it a condition of publication that studies are conducted in accordance with the Declaration of Helsinki. How the guidance offered by the Declaration should be implemented in practice still causes some debate. Recently this focussed on the design of clinical trials where a proven alternative exists to the new treatment under investigation. Rothman, Michels & Baum (2000) argued that the new treatment should be compared with the proven alternative and not with placebo and that the use of a placebo group in this particular situation is unethical. At their meeting in Scotland in 2000, the World Medical Association adopted this view which is now stated explicitly as principle No. 32 [see and B-3]. For further information on ethical considerations in clinical trials see sections B-1, B-3 and B-8 of this handbook.

Back to top

F-4 Other research issues

F-4.1 Research governance

All research should be carried out to high scientific and ethical standards. Research governance is the process by which these are ensured. Sponsors and funders of research are expected to conform to these standards and to have procedures in place to ensure that research projects for which they are responsible do, too.

F-4.2 Data protection

All research on human subjects should be done under data protection principles, ensuring the confidentiality of personal data and restricting the use of data to applications for which consent has been given. In the UK, this is the responsibility of the Information Commissioner. Applications should include a statement as to how data protection principles and legislation will be complied with.

Back to top

References for this chapter

Altman DG (1982) Misuse of Statistics is unethical. In Statistics in Practice (Eds SM Gore and DG Altman). British Medical Association, London.

Altman DG. (1991) Practical Statistics for Medical Research. Chapman and Hall, London.

Rothman KJ, Michels KB, Baum M. (2000) For and against. Declaration of Helsinki should be strengthened. British Medical Journal 321 442-445.

Back to top

Back to Brief Table of Contents.

Back to Martin Bland's home page.

This page is maintained by Martin Bland.

Last updated: 6 January, 2012.

Back to top