Good data management in social science research is of crucial importance. Recent fraud cases definitely helped to increase awareness of this among researchers. But is fraud detection, or: fraud prevention, the main reason why we need good data management? Of course, good data management can help to detect fraud in an earlier stage and it may even help to prevent it in some cases. But there is more… Decisions at different stages Research questions and designs in social science research have become increasingly complex. More and more studies require complex decisions at multiple stages and different people may be involved in different decisions. International research projects are quite common these days. Questionnaires and other materials are translated from one language to another, and it has surprised me that in quite a few studies that I am aware of there is a very minimal check – if any – on the quality of translations. In any stage when translation from one language to another occurs, at least two people who are proficient in both languages should be involved, and translations (and perhaps potential alternatives) should be archived. The same holds for professional language. Experts in a domain or methodological tradition use language that most outsiders are not familiar with. Whether a study involves qualitative methods, quantitative methods, or both, decision making is a subjective process. Even in seemingly ‘simple’ studies, multiple approaches to analysis may exist. Assumptions and choices – at any stage – should be archived, and choices made should be backed by at least two experts. Swiss cheese… or… any cheese? As a statistician, I sometimes receive a dataset that looks somewhat like a Swiss cheese, with many holes including big ones, along with the question from the principal researcher(s) what – if anything – we can still get out of that dataset. Responses to my follow-up question why there is so much data missing include ‘we lost part of the data due to a [computer/server] crash’ and ‘an email folder accidentally got erased’. I thought that this was about as bad as things could get, until I got involved in the following. Some of my colleagues wanted to take a look at a range of studies focusing on a particular phenomenon and see how we could summarize these studies and inform practice and/or future research. This project involved having to contact authors of articles presenting one of the studies under consideration with the question to provide specific information with regard to for instance an average score or standard deviation. We sent out emails to quite a number of authors from whom we needed more information in order to include their studies in our project, because their articles did not provide specific information that we needed for our project. More than once, we received an email stating something like ‘I moved from one department to another and lost all data’ or ‘the principle researcher no longer works here and the data went with [him/her], too’. Let me add that the studies under consideration were published in the last say four to seven years. I cannot help it, but I have to be honest and admit that such responses made me wonder whether we are really living in the 21st century. Good data management serves more than one purpose The examples that I have provided appear to point out a certain sloppiness rather than fraud and in some cases just subjective decision making (translation and decisions by experts). I think that fraud represents just a very small part of social science research practice, as is the case in virtually any kind of profession we can think of. The fact that some members of a professional community display fraudulent behavior does not imply that this is common practice in the community. Of course, good data management can provide a good tool against fraud. However, I think that it is even more important that it provides a safeguard against a certain sloppiness in research. We are living in times when research has to be carried out and published under a certain time pressure. Moreover, many researchers have to balance their working time between research and a variety of other activities. In such a context, it is at times easy to overlook something, to make a hasty decision, and to perhaps even forget to make a backup while in the process of collecting data. To help researchers in such a context, data management must be a fundamental part of research and research philosophy of the institution or department in which the research is carried out. Networks like Dataverse (thedata.org/) provide great online environments for the storage of research data and enable researchers to explain who have been involved in the research and what choices were made at different stages. Apart from all discussion about fraud and sloppiness – a discussion that has reached all newspapers and platforms – good data management simply enables us to strive for continuous development and improvement of the kind of research we are doing. Research is never perfect; looking back at research carried out, we always realize things we did not realize before, and there are always things we would do somewhat differently if we could do the same study again. The great thing about good data management is that anyone interested can learn from choices that did or did not work out well, and this may contribute to better and better research that has more and clearer implications for the practice we wish to inform.
Posted on: Wed, 21 Jan 2015 00:14:47 +0000
Trending Topics
Recently Viewed Topics
© 2015