Peerwith connects researchers with with peers who are experts in researcher services. These service go well beyond language and editing. There are also experts on Peerwith that provide services in for example Data and statistics. Today we introduce you to Tomoe Gusberti, expert in that field. We asked Tomoe to express her perceptions on the current research panorama, positioning the value of statistic and data analysis services in the collaborative peer-to-peer model unique for Peerwith.
How to take the most from the collaboration with data and statistical analysis experts: What you must know about the professionalization of research and knowledge generation industry
The research and knowledge generation industry is professionalized, competitive and aligned with the so-called knowledge economy. Transdisciplinarity, complexity, and sophistication in study conduction, analysis, and diffusion of the resultant knowledge are argued as an undeniable reality.
There is a trend toward specialization of the tasks related to the production (study design, statistical and data analysis), storage (data and information management), and dissemination (paper development, publishing, knowledge/technology transference) of the knowledge 1–4.
|Relevant universities provide support services as statistical analysis and data management system. But this structure is becoming big, complex, and no all universities and research institutions can offer and keep this kind of structure internally. Hence, although initially outsourcing practices were focused on non-core activities, Higher education is now outsourcing core-activities5|
It is not because your institution does not offer such support it is acceptable to submit or deliver useless and non-relevant research results/papers. The research and knowledge generation industry is competitive. To worth the funding, scholarship, consent enabled by participants of your research, and other shareholders and investors of the infrastructure you are using for conducting the research, you must do your best to get relevant, reliable, valid and reproducible research results. In the current and practical terms, it means: publish in indexed journals. Luckily we have Peerwith!
Effectiveoutsourcing is not just about the availability of professionals or companies offering language edition or statistical analysis services. Outsourcing (generally speaking), if not well managed, present struggle in matching the more suited professional, difficulties in communication, lack in philosophical and strategical alignment between the parts5, and a risk to compromise the quality6.
The peer-to-peer model at Peerwith provides not just a way for outsourcing. Some important research support services are not as commodities, and communication and collaboration are essential. Common pitfalls in academic research and papers writing are the lack of alignment with the theory and adequate statistical analysis. This post discusses Peerwith’s statistical review and analysis services.
In essence, Statistical review is a specialized modality of pre-submission peer review focused on the contents comprising study design, data collection methods, statistical analysis and interpretation of the results. Methodological issues are one of the most common pitfalls that make the manuscript away from publishing. If your manuscript comprises a statistical analysis, certainly better to make clear you need an expert with a statistical background, selecting the “statistical review” category.
Requesting a “statistical review” is a way to specify that you are requiring the technical review tobe focused onstatistics/data analysis. You also must be aware that:
- Usually, an output for this kind of service is a report in more recommendations, line, not including direct modifications in the text;
- If the expert considers the need for complementary analysis or generate new summarization tables and graphs, for example, the suggestion to conduct such analysis is one result, but its conduction is not included;
- It usually does not include load again or refining your analysis. Some journals mention they require their referees to reload the analysis, but not all journals… you can think it is advisable to anticipate this;
- The reanalysis of the new improved version of the manuscript after implementing recommendations is usually considered a new request.
It is a good practice to specify in your request:
· which level of service you want, just the statistical part, or also all the other issues;
the deepness of the review:
o if you expect modification in the text, or just punctual and general comments
o If you believe it is important to the expert reload the analysis, you must specify how (SPSS/Minitab/R/Matlab/SAS…) if you are applying complex methods that can present some small differences in algorithms implementation or outputs availability, as you cannot expect all the expert having access to all proprietary software;
· If you wish the expert being available for further discussion for one week to enable you to carefully read and consider for implementation of the recommendations, and ask for more details about the recommendations, it is a good idea to specify this.
After a statistical review, the modifications recommended by the expert can be simple to adjust, but, can also show the need for significant improvement. The most probable negative response is not about an error in calculation (because everybody uses software). There is a used expression: “who have an only a hammer see all stuff looks like a nail”7. If your statistical/analytical background is limited than the current practice in the research field, better to count on the support of some expert with statistical/data analytical background. It is possible the method you know is not suited for the data you collected or the specific objectives of your analysis.
It is not just about simplicity x complexity or fashionable method, although we know some research fields and journals has implicit (sometimes explicit) preferences for some methodological approaches and techniques. And yes, a negative response because “analytical method is somewhat simplistic” happens, but usually is not just about that.
As journals referee, I received papers applying simple and well-known basic methods as t-test, ANOVA, OLS regression, ignoring the associated assumptions as normality of data/error, homoscedasticity for example. When the method shows inadequate all result and discussion become useless and, as referee, I have no option but to recommend “revise and resubmit” or sometimes even a clear “reject”. It is hard to wait for a journal editor or referee look beyond the data and analytical results you are presenting in the manuscript. If the potential rescue is clear, referees can try to recommend a “review”, but you cannot expect this because this is not foreseeable with just summarizations about data structure and composition in a manuscript.
Well, with the statistical review service, you do not lose 3 months waiting for this response (usually at most 3 days is sufficient for this kind of review), but you could avoid this if you collaborated with professional with data and statistical background previously.
If you have an idea of your analytical plan, you can request in “Study design” a verification to check if your “hammer” suits for the specific case you are dealing… but you also can request a “statistical and data analysis”.
How far we can expect for the partnership and collaboration with an expert?
If you just join your literature review, with the statistical analysis report, it will not be an academic report/paper/thesis. An academic report requires further discussion, joining and transitioning the theoretical background of the applied knowledge field, and statistics/analytic/methodological considerations.
Do not let your paper or thesis be a Frankenstein.
A good statistical/data analysis project will emerge if you consider the interaction with the expert as actual collaboration, making the expert your collaborator, partner in the research project, co-authoring the paper with you. It is not just about ethicsor coping with COPE’s or other Publication Ethics Guidelines followed and required by publishers. It makes collaborationmore fluid. A partner indicates pitfalls, challenge your rationales, interpretations, indicate limitations and opportunities to be suggested as recommendations for future research, and will just satisfy when the manuscript matches the quality standard that worth his/her work, and would be acceptable for publication.
The paper with a deep discussion will emerge from all these convoluted discussions:
- Let the interaction with the expert be a transdisciplinary discussion.
- Make discussionsquestion the expert about your doubts;
- take the opportunity to learn, listen/read carefully the critics and pitfalls about the data already collected.
- It can sound a little unpleasant sometimes, but knowing research limitations, and possible bias enable to make pondered considerations
Those pondered considerations are an actual important task in research and decision making.
Will the expert write the results and discussion section of the manuscript?
Being direct, No. You cannot expect this. To promise writing the results in the manuscript style is one thing, but the results and discussion section for all research paper is another thing. This cannot emerge just from the statistical or data analysis, but from the transdisciplinary discussion.
Hence, I will never promise fullresults and discussion sections… this avoidsmisinterpretation, assure that the main author deal with the results and understands the analysis, and also enable a better evolution in the results interpretation… It is also a way to avoid failing in the several shades of ghost writership and other unethical behaviours we cannot agree, and also stronglycontested bythe relevant and high impact academic journals…
In writing the report with several extravagant comment frames, I direct questions and discussions with the client to make the results interpretation evolve… the erudition of the analysis comprises two items: (i) the sophistication in the analytical method; and (ii) the deepness in the discussion of the results, connecting the analytical results, the theory, and the contextual issues, including implications for generalizations, research limitation and opportunities for further research.
I would like to reinforce: the collaboration is a key for good transdisciplinaryresearch dealing with complex problems, an objectof studies and phenomena. Do not let your paper or thesis be a Frankenstein. I try the most to make the discussion to flow as much as if we were side by side, even with several hours of time zone differences.
- Kennan, M. A., Corrall, S. & Afzal, W. “Making space” in practice and education: Research support services in academic libraries. Libr. Manag.35, 666–683 (2014).
- Snyder, D. C. et al.Research Management Team (RMT): A Model for Research Support -Services at Duke University. Clin. Transl. Sci.5, 464–469 (2012).
- Bryant, R., Lavoie, B. & Malpas, C. The realities of Research Data Management: Selected University RDM Service Profiles. (OCLC Research, 2017). doi:10.25333/C3T92S
- French, J. & Poznanski, D. Managing Research-A Guide to Research Infrastructure. J. Med. Imaging Radiat. Sci.40, 165–169 (2009).
- Insinga, R. C. & Werle, M. J. Linking outsourcing to business strategy. Acad. Manag. Perspect.14, 58–70 (2011).
- Wekullo, C. S. Outsourcing in higher education: the known and unknown about the practice. J. High. Educ. Policy Manag.39, 453–468 (2017).
- Breiman, L. Statistical Modeling: The Two Cultures (with comments and a rejoinder by the author). Stat. Sci.16, 199–231 (2001).