Data publication: “Part & parcel” of the reproducible research process

Joost de Win­ter is an asso­ciate pro­fes­sor work­ing in the Depart­ment of Cog­ni­tive Robot­ics at the Fac­ul­ty of Mechan­i­cal, Mar­itime and Mate­ri­als Engi­neer­ing (3mE) at the TU Delft. Joost’s work focus­es on Human-Robot inter­ac­tions specif­i­cal­ly in the field of auto­mat­ed cars. 

“If you hear the name of the dis­ci­pline Mechan­i­cal Engi­neer­ing maybe you don’t imme­di­ate­ly think about data, but in the sec­tion where I work it is all about the data. I work on the top­ic of human inter­ac­tion with auto­mat­ed cars where we col­lect dif­fer­ent types of data, for exam­ple, record­ed data from sen­sors, mod­els, and algo­rithms on how to process the data.”

Joost de Win­ter

In recent years, Joost has worked inten­sive­ly to imple­ment best prac­tices for data shar­ing with­in his group and to raise aware­ness about the impor­tant role of data in ensur­ing repro­ducible research. His inter­est in repro­ducibil­i­ty in his own field of research moti­vat­ed Joost to apply for a repli­ca­tion study about ‘pupil size as relat­ed to inter­est val­ue of visu­al stim­uli’ financed by the Nether­lands Organ­i­sa­tion for Sci­en­tif­ic Research (NWO), where he also col­lect­ed valu­able expe­ri­ence on the gen­er­a­tion of bet­ter and repro­ducible research. 

“Five years ago I start­ed to real­ize that we need to be more care­ful with the data. When read­ing about all the repro­ducibil­i­ty prob­lems in psy­chol­o­gy, for exam­ple, I thought that we could improve our work­flows as well, from the col­lec­tion to the pub­li­ca­tion of data.”

To achieve this, Joost estab­lished a set of prin­ci­ples for his group mem­bers on where to store the data to make it avail­able for the whole group as well as how to orga­nize and struc­ture fold­ers.

Joost and his group have archived a col­lec­tion of data sets on 4TU.ResearchData. His main moti­va­tions to pub­lish data are ensur­ing repro­ducibil­i­ty and account­abil­i­ty of the research con­duct­ed by his group. 

“If you pub­lish some­thing, you want to ensure that your research is reli­able. As a min­i­mum our group wants the results of our papers to be repro­ducible. The data used to cre­ate fig­ures and all the analy­sis done have to be avail­able for inter­nal review as well. When you make the data avail­able (as much as pos­si­ble) it gives more con­fi­dence in the study.”

The rapid advance­ment of research on automat­ing dri­ving is result­ing in the col­lec­tion and pub­li­ca­tion of large amounts of data, bring­ing huge poten­tial for new research. Joost and his group will con­tin­ue con­tribut­ing to the field pub­lish­ing their data on 4TU.ResearchData.

“The pub­li­ca­tion of datasets is very easy and straight­for­ward. 4TU.ResearchData pro­vides per­son­al sup­port and cus­tomized help, the data are in-house (insti­tu­tion­al servers), it com­plies with all require­ments for a stan­dard­ized repos­i­to­ry and it’s free!”

At the end of our con­ver­sa­tion, we asked Joost what he would tell oth­er researchers to encour­age them to pub­lish their data. His response was:

“First of all, the research process is only ful­ly com­plet­ed if you can repro­duce the study. There­fore, the pub­li­ca­tion of data is part of the research process. If there is a need for more encour­age­ment, I can say that pub­lish­ing your data ensures the con­ti­nu­ity of projects (sub­se­quent stu­dents have access to the data). If some­one rais­es ques­tions about your research they can go direct­ly to the repos­i­to­ry and check the data, mul­ti­ple peo­ple can go through the data and check for mis­takes and it gives you a good feel­ing to final­ize the project hav­ing all the data com­piled and find­able.”

All of the pub­lished data sets gen­er­at­ed by Joost de Win­ter and his group are avail­able here.
Cov­er image by Chaitawat Pawapoowadon via Pix­abay