When submitting a paper to a highly profiled publication, like Nature or Science, author contributions must be clarified. This may include study design; data acquisition, analysis, and interpretation; draft preparation and/or critical editing and review of the manuscript. Among these aspects, I was wondering how data analysis and data interpretation are different? Consider the boundaries may not be quit clear, because more often it is difficult to analyze data to find true patterns without being able to interpret those patterns.
How contribution in "data analysis" and "data interpretation" are different, for a scientific paper?
3
-
I would think that data analysis would mean the author crunched some numbers, whereas data interpretation would mean the author looked at the output graphs and gave insightful comments. I may be mistake, but I would think that's common use.– Chris RackauckasCommented Apr 24, 2016 at 4:14
-
Students do data analysis: getting their hands dirty with time consuming computer codes for statistical analyses. Professors do data interpretation: looking at graphs and statistical quantities once they've been calculated by a student ;)– innisfreeCommented Apr 24, 2016 at 5:11
-
@innisfree it is a little bit difficult to understand your comment when it is a Nature paper ;)– AvestanCommented Apr 24, 2016 at 5:33
Add a comment
|
3 Answers
I agree that the boundaries are not quite clear in that it is impossible to analyse a set of data without giving some sort of interpretation to them at the same time.
The requirement however doesn't mean that you have to identify a single author for each contribution. For example, you can say that author A contributed to data acquisition, author B to both data analysis and interpretation.
In an experiment, data acquisition can simply mean running the tests without processing the data into more meaningful plots or graphs. The latter falls into the category of data analysis. Discussing what the data mean comes into interpretation. While interpretation cannot be absent in any analysis, and so I cannot imagine an author contributing to analysis but not to interpretation, the reverse may be possible. An author can contribute to interpretation without contributing to analysis.
-
so in case one (myself) is among the data analysts but believes should be acknowledged for data interpretation as well, do you have any comment on how to politely discuss this with authors in a multidisciplinary group from different institutes?– AvestanCommented Apr 24, 2016 at 16:29
-
I have no experience with this. I suggest that you post this as a separate question?– adiproCommented Apr 25, 2016 at 6:15
I'm frequently on the "data analysis" end of this spectrum when I work together with biological psychologists (or psychological biologists). Basically, I run the analyses, do all kinds of diagnostic checks, plot enlightening plots and discuss all this with the biologists and the psychologists. They in turn look at my plots and know what results make biological/psychological sense, e.g., which biomarkers are frequently associated with what disease or other, or what the recent literature has to say about some specific relationship. Or, much more basic, the biologist can tell me which measurement makes sense, and which one has to be a measurement error.
This is the relationship between data analysis and data interpretation.
I think that this kind of division of labor makes a lot of sense. Statisticians simply know a lot more about statistics than do psychologists, doctors or biologists. And (sorry) I have seen very disheartening things happen if subject matter experts think they know enough statistics to run complicated analyses themselves. Yes, some psychologists do have an excellent grasp of statistics - but most simply don't. And that's how it should be. After all, I don't expect my car mechanic to be able to repair my refrigerator, either. (It should go without saying that I don't think statisticians shouldn't arrogate any specific subject matter expertise to themselves, either.)
The relationship between subject matter expertise and statistical knowledge is a frequently discussed topic in what is nowadays called Data Science. I have expounded on my point of view here.
answered Apr 24, 2016 at 21:30
-
I've upvoted this because it was helpful to me (philosophy PhD plus back in the day Chemistry BS). Do you think the distinction is always this clear? What if the analysis methods used are not so difficult? My sense before reading this is that the data interpretation could include lots of comparisons with current literature that helps to explain what the numbers resolved in the analysis mean.– virmaiorCommented Apr 25, 2016 at 10:27
-
@virmaior: no, it's not always this clear. Sometimes there is, in fact, one person who is competent in both data analysis and interpretation, or some subject matter expert may know enough data analysis tools and statistics to do his own analysis and interpret it. However, I have a bit of a suspicion that there are few low-hanging fruit left in many natural sciences, i.e., analyses that can usefully be run without advanced understanding of statistics. Most large effects have been found, and to tease out the smaller ones, simple tools are often not enough. Commented Apr 25, 2016 at 11:29
-
@virmaior: and yes, knowledge of the subject matter literature is exactly what the domain experts contribute. I find it hard enough to keep up with the statistical literature - there is no way I could also follow the psychological or the biological literature. (Nor could the psychologists or biologists. Although it makes sense, of course, to recommend carefully selected articles to one's collaborators whose expertise lies in different fields.) Commented Apr 25, 2016 at 11:31
You should certainly be able to do both. Data analysis incorporates data management and statistical analysis. In data interpretation you find out what the data means, how they throw light on your research questions. Data analysis is necessary for interpretation; however, data analysis is also guided by your anticipations, i.e. your research questions which again, hopefully, will relate to what the data will tell you (because you designed your data sampling so).
