Is That Clear?
Novie Younger, Tropical Medicine Research Institute, University of the West Indies
Novie Younger works at Tropical Medicine Research Institute at the University of the West Indies in Jamaica. Her research interests include analysis of complex survey data, survey methodology and data quality, teaching statistics in the pure and applied sciences, reliability and validity of screening tools, and analysis data from epidemiologic studies.
Heavy use of statistical jargon can be a deterrent to the application of statistical inference or appropriate use of results. As statisticians, the make-up of our clientele ranges from students to collaborators in research to media personalities and lay people trying to make sense of the results we produce. How can we ensure our results move beyond high-impact, peer-reviewed journals to active use in industry and everyday life? It begins with making our results more lucid to those with little exposure to statistics.
Advice for Beginning Consultants
In a consulting relationship, statisticians need to lay ‘ground rules’ early on for coauthorship, consultation length, and remuneration. It is important that clients are clear about our expectation of coauthorship so our efforts do not remain unrecognized. This can be done if the relationship between client and statistician is built in a manner that minimizes difficulties associated with our becoming coauthors. In academic settings, for example, some statisticians waive payment for services rendered in exchange for coauthorship with graduate students or faculty.
It is also important that we apprise clients of the period that will be dedicated to direct consultation or collaboration in preparation of manuscripts for publication. The acceptable remuneration for services rendered must also be discussed and agreed to early in the consultation. Profitable discussion regarding remuneration will make our clientele aware of the work involved in data analysis. Clients should know that data management and model adequacy checks are important aspects of data analysis and help yield the best results.
It has been said that “research not published” is “research not done.” It is also thought better to leave a research question unanswered than to carry out analysis that answers the question and leaves the results unused or, worse, hidden from publication. Thus, in the early stages of consultation, we should inquire about the utility of the results if they are to fall in either extreme of possible findings. If the extremes are not useful, the collaborator should be encouraged to reconsider doing the project. Often, decisions that have to be made based on the results are subject to managerial and budgetary constraints, rather than statistically significant differences. Nevertheless, analysis carried out can illustrate the best- and worst-case scenarios, even if there cannot be direct implementation of findings.
During the 2008 Joint Statistical Meetings, I led a roundtable at which we explored methods for effectively communicating results and the use of statistical methods to people who need to use statistical tools and apply research results, but do not have a strong numerate background. The discussion made us recognize that appropriate use of results and methodology by practitioners outside our profession can lead to more discussions of data analysis results in everyday life. One of the early steps we can take is to train students in consulting, so they will begin to learn the value of giving statistical advice in a clear manner before they begin to practice formally.
As statisticians, we have a tendency to think there are certain facts everybody should know. However, our experiences show us there are basic competencies our clientele may not have. These include knowledge of how a data set should be laid out and features of a data set that will ensure successful and useful data set merging. Thus, to facilitate effective communication between statistician and client, it may become necessary to provide a document that outlines basic methodology relevant to data manipulation.
It is important that we become involved in training those learning the applied sciences or working in applied settings to become conversant with statistical methodology. We could host and attend seminars, brown-bag lunches, and journal clubs at which we learn each other’s languages. At all levels of the institutions in which we work, we should encourage statistical thinking. We should encourage researchers and academics to publish data analysis findings that answer even simple research questions. Such publications, written in a manner readily understood by those who really need to use the findings, will more readily lead to the widespread use of statistics.
There are times when lack of interest in a project or research question leads to impaired and ineffective collaboration. It is useful to encourage other specialists in the research team (e.g., public health specialists or epidemiologists) to develop their own research projects or questions that can be answered using data from an existing project. To heighten interest, it may even be necessary to modify a main research question or objective.
What can we do to promote uptake of data analysis findings by relevant stakeholders? We can encourage our clientele and collaborators to practice their presentations in front of us prior to presenting them in a more public domain. In addition, we should feel privileged to see manuscripts based on our work prior to their submission for publication. And, when possible, we should accompany collaborators to the meetings at which they present findings based on our data analysis.
Use of complex statistical methodology for data analysis often produces more robust results that, if used to drive policy decisions, could lead to more appropriately targeted interventions. Unfortunately, such results do not go beyond the journal in which they are published at times, leaving human welfare deprived of further development. We must make intense efforts to convert our findings to language that stakeholders can apply. This can be aided by the use of short, simple sentences and appropriate, easily understood pictures. It is also important that we simplify results for newspapers, newsletters, and the general media. We have to be especially mindful of whether our collaborators can explain results to the media in a clear and correct manner.
As statisticians, we carry out highly valued work that can greatly improve human welfare. Let us make every effort to maximize our efforts through effective collaboration and education.