FEATURE

What to expect when meeting a

BY G CZANNER

here are a growing number of “To consult the statistician after an The typical questions are: working closely with is finished is often merely • What are your research questions? ophthalmologists. They have to ask him to conduct a postmortem • Have you already collected the ? Tdifferent training but they are examination. He can perhaps say what the • If you have already collected the driven by the same goal: to perform high experiment died of.” [4] data, what is the study type (e.g. quality evidence based clinical research retrospective observational or case / [1,2]. In a perfect world we would simply Skills and education control study)? conduct clinical studies on all patients Statistical consultants have postgraduate • What is the study population (e.g. and observe differences between the training in (MSc or PhD), as age range, control groups, inclusion / treatment groups, with the conclusion well as relevant practical experience, exclusion criteria)? that the observed differences are due which may be formally recognised via a • What are the outcome measures to treatment. However, due to ethical statistical society (e.g. CStat is a chartered (primary and secondary outcomes, reasons, we can only conduct a study statistician recognition given by the including the type of data that are on a sample of patients, which brings Royal Statistical Society and currently being collected, for example binary, uncertainty into clinical research. In re-evaluated every five years). Advisors categorical, continuous, time to particular, we need to decide if the may also have significant education and event)? variability observed is due to true experience in the particular field in which • What is your funding? differences between treatments or simply they work (e.g. medicine, neuroscience, due to chance. This is one of economics, manufacturing). Some What range of tasks can a the questions that can be answered by universities offer specific graduate statistician do? statistics and was well characterised by programs in fields such as , The type of tasks that a statistician can my statistics consulting teacher Allan psychological methods, or methodology do vary depending on funding and on Sampson (University of Pittsburgh): and statistics for the medical, behavioural the specifics of the project. A statistician “Statistics is the language of science; and social sciences. can simply give advice or can also do the in the omnipresence of randomness, analyses. The statistician may just do a statistics provides the context and Where do we find statisticians? single consultation at one point during the framework for the understanding and Statisticians work in universities, in study (called cross-sectional consulting), interpretation of data.” industry, consultancy or as self-employed they may repeatedly give advice or free-lance consultants. Some universities do some analyses thoroughout the Statisticians and statistical offer in-house statistical advice for duration of the study (called longitudinal consultants researchers from within the university, consulting) [3]. Typically, a statistician can A statistician is a person who can or they may have an established self- provide advice or assistance with: perform analyses, give consultations, financed consultancy centre. Statistical • Formulation of the hypotheses. A teach and perform research to invent advisors are also consulted by public research aim needs to be formulated new statistical methods. A statistical administration, where they may be into testable hypotheses. This usually methodological advisor or consultant involved at all levels of governance. In the starts with a discussion of what the provides methodological and statistical legal system, statistical consultants may primary and secondary outcomes need advice and guidance to clinicians be called upon as expert witnesses, in to be. interested in making decisions regarding particular in cases that involve statistical • Choice of study design and research the design of studies, the collection and considerations. study (e.g. case-control study). analysis of data, and the presentation • Determination of an adequate and dissemination of research findings What type of questions will a sampling procedure for your study, [3]. Trained in both statistics and statistician ask? survey or experiment. communication skills, advisors can work Typically, a statistician will want to • Randomisation for randomised clinical in academia, industry or the public sector. know what type of advice is needed, the trials. research questions, the resources for • Determination of sample size When is the best time to seek statistical work and current progress with calculations (i.e. number of patients in statistical advice? data collection or analysis. Professional each treatment group). An honest answer is to seek statistical statisticians and consultancies will ask the • Choice or construction of advice before beginning the research. client to fill in a questionnaire or they will measurement instruments. To quote statistician Ronald Fisher: fill it in at the first meeting with the client. • Data analysis methods (e.g. t-test,

eye news | AUGUST/SEPTEMBER 2014 | VOL 21 NO 2 | www.eyenews.uk.com FEATURE

, logistic • Presenting your collected data (e.g. substantive research questions and to regression, etc.) so that the research advice on best plots). translate them into methodological and questions can be answered statistical procedures. In some cases, adequately. Simple vs. advanced statistical this may also entail profound ethical • Systematic review and meta-analysis. methods considerations [3]. • Statistical advice or participation in Sometimes only advice is sought from a References a grant proposal. It is necessary to statistician with the understanding that 1. Bunce C, Patel KV, Xing W, et al. Ophthalmic address all of the above areas in a the clinician will do the analyses. In such statistics note 2: absence of evidence is no grant proposal. The statistician can cases, the statistician needs to know evidence of absence. Br J Ophthalmology 2014;98:703-5. give advice, write the grant proposal the clinician’s level of understanding 2. Holopigian K, Bach M. A primer on common section and estimate the funding of research methods and statistics. If statistical errors in clinical ophthalmology. Doc needed to do any data analysis work. their understanding is small then the Opthalmol 2010;121:215-22. 3. Ader HJ, Mellenbergh GJ and Hand DJ. Advising • Supervision of data collection to statistician should suggest both a simpler on research methods: A consultant’s companion. ensure elements of the population are statistical method and more advanced Huizen, The Netherlands; Johannes van Kessel Publishing: 2008. being sampled correctly. methods. 4. Fisher RA. Presidential address to the First Indian • Presentation of results, help with Statistical Congress. Sankhya 1938;4:14-7. write-up of statistical results for Defining the statistical manuscripts, professional conferences hypothesis and slide presentations. A statistical consultant will help to clarify

Gabriela Czanner, PhD CStat, TAKE HOME MESSAGE Department of Biostatistics, Faculty of Health and Life • It is important to think about study design and data analysis methods at a Sciences, Department of very early stage of clinical research, and it is crucial to have discussion with a Eye and Vision Science, Faculty of Health and Life statistician who can help to formulate the hypotheses. Sciences, University of Liverpool; • Statistics contains a very large number of methods and tools; and often Declaration of Biostatistician, Clinical Eye Competing Interests: statisticians specialise in some areas, while striving to have good knowledge Research Centre, St. Paul’s None declared. and experience in common data analysis methods. Eye Unit, Royal Liverpool and Broadgreen University • Good communication and trust between statistician and ophthalmologist is Hospitals NHS Trust. one of the most important keys for successful clinical research. E: [email protected]

eye news | AUGUST/SEPTEMBER 2014 | VOL 21 NO 2 | www.eyenews.uk.com