By: Henry Walker

Henry Walker — Tue, 10 Jul 2007 18:17:48 +0000

Dr Hollister: With a bit more participation this could become a very interesting Blog. I’m sure there are interesting examples of Ecological Informatics and Reproducible Research out there.

In addition to KNB, EML, Kepler and Analytical Webs, there is the basic business of referencing and sharing data.

I’ve recently come accoss “Dataverse Network” G. King 2007. who points out the possibility of citing the data used in a paper using a Universal Numeric Fingerprint (UNF). The R package UNF by Micah Altman computes a UNF based on the data. People can then search for the UNF if they want to obtain the identical data set used in a paper. even if it moves to a different URL. After downloading a dataset, one can regenerate the UNF to be sure it is the same data used in a paper.

Recent open source R packages also help make it possible to share data, script, and even cashed R computations: Eckel and Peng (2006), Peng (2007), Peng and Eckel (2007)

Approaches involving UNFs for data, the use of open source script, and cashed computations for the analysis, and reproduction of published figures and tables will facilitate :
(1) more critical peer review of research results,
(2) technology transfer of computation methods that others can adapt, and
(3) a shift from advertisment or research results and advocacy arguments over alternative interpretations data, to quantitative weight-of-evidence approaches based on Information Theory , Burnham and Anderson (2002).

We can make more credible advances in ecology and ecoinformatincs, based on (1) more effective data sharing, and (2) the adoption of reproducible research approaches.

References:

Gary King 2007 Dataverse Network. http://gking.harvard.edu/talks/dvn-nsfP.pdf & @ http://thedata.org/index.html

Sandrah P. Eckel & Roger Peng 2006
INTERACTING WITH LOCAL AND REMOTE DATA RESPOSITORIES USING THE stashR PACKAGE. John Hopkins Working Paper 127. http://www.bepress.com/cgi/viewcontent.cgi?article=1127&context=jhubiostat

Roger Peng 2007. A REPRODUCIBLE RESEARCH TOOLKIT
FOR R. John Hopkins Working Paper 142. A REPRODUCIBLE RESEARCH TOOLKIT FOR R

Roger Peng and Sandrah P. Eckel 2007. DISTRIBUTED REPRODUCIBLE RESEARCH USING CACHED
COMPUTATIONS John Hopkins Working Paper 147. http://www.bepress.com/cgi/viewcontent.cgi?article=1148&context=jhubiostat

Burnham and Anderson (2002) Model Selection and Multi-Model Inference. Springer.

————————————————-

Henry A. Walker, PhD
EPA ORD NHEERL Atlantic Ecology Division
Narragansett, R.I. 02882

By: Henry Walker

Henry Walker — Tue, 03 Jul 2007 14:00:08 +0000

Scientists in diverse fields of study are increasing recognizing the value of transparency, reproducibility. By making data available and computational methods transparent and reproducible, one can anticipate more rapid research advancements. It makes it possible for peer reviewers to examine the data and computations in more detail during the review process, reducing the likelihood of publication of fraudulent results, a journal’s worst nightmare (Laine et al, 2007). The approach is gaining traction in the biomedical field. In addition to researchers learning the practice, for the approach to succeed more broadly major journals will need to embrace it.

Henry A. Walker U.S. EPA, National Health and Ecological Effects Research Laboratory, Atlantic Ecology Division, Narragansett, R.I.

Christine Laine, et al 2007. Reproducible Research: Moving toward Research the Public Can Really Trust. Annals of Internal Medicine. Volume 146 Issue 6 | Pages 450-453 http://www.annals.org/cgi/content/full/146/6/450.

Comments on: Moving Forward with Ecological Informatics and Reproducibility

By: Henry Walker

By: Henry Walker