Bibliometric analysis of Danube region output. A. Degelsegger / D. Lampert, ZSI Danube-INCO.NET GA workshops Vienna,

Transcription

1 Bibliometric analysis of Danube region output A. Degelsegger / D. Lampert, ZSI Danube-INCO.NET GA workshops Vienna,

2 Outline Presentation (1h): Context of the study Methodology Preliminary results (extracts) Break and results exploration (0.5h) Group discussions (0.5h) Plenary discussions (0.5h): Reporting back General discussion Relevant next steps

3 Presentation of the study methodology? 3

4 Presentation of the study Conduction of the study process: methodology 4

5 Sources (WoS, Scopus, etc.) Presentation of the study methodology import Raw data correction Records (corr./normalised) SQL queries Results tree (description/csv/sql) selection, inter pretation, contextuali sation Conduction of the study process simplified version Presentation (reports, visualisations) 5

6 Presentation of the study methodology Conduction of the study process: 1. Definition of the data set(s) 2. Data Retrieval 3. Data cleaning/normalisation & processing 4. Analysis 5. Visualisation 6. Reporting 7. Presentation with Feedback; Report-Finalisation 6

7 Language record = a publication (article, letter, etc.) indexed in Scopus or Web of Science co-publication = internationally co-authored indexed record impact = impact in the research community; citation counts are used as a proxy Danube region as defined in Danube-INCO-NET; for this presentation, taking into account the geographical distribution of the participants, the following countries have been in the focus: AT, BG, CZ, DE, BiH, HR, HU, MD, ME, RO, RS, SK, SI, UA STOP US IF NECESSARY

8 Presentation of the study methodology Data Retrieval Thomson-Reuters Web of Science Elsevier Scopus retrieved in summer 2014 Error types Missing information author first name, no affiliation address Spelling mistakes Wien / Wein Variable spelling Łódź / Lódz / Lodz / Copy mistakes Austria instead of Australia 8

9 Presentation of the study methodology Import of raw data Duplicateidentification and correction Record examination Affiliation data unification Data unification Classification (journals) Manual correction of countries and addresses Decomposition of organisationstrings Organisationclustering on two levels Data cleaning/normalisation & processing workflow 9

10 Presentation of the study methodology Data cleaning/normalisation & processing Import BibTeX-files (WoS, Scopus) selection and inclusion of relevant data chunks monitoring of import success new data update older data in records 10

11 Presentation of the study methodology Data cleaning/normalisation & processing Duplicate Correction Using identifiers of WoS and Scopus Multiple imports (keep times-cited counts up-to-date) Remaining duplicates: auto-matching with identifier db_source DOI author title Year Volume pages Manual approval of ambiguous candidates to be merged 11

12 Presentation of the study methodology Data cleaning/normalisation & processing Raw Data Correction Automatic error detection wrong DOIs, ISSNs, ISBNs ambiguities during parsing process missing fields invalid information Raw Data Correction: > 5000 records with errors ~ 900 fixed manually 12

13 Presentation of the study methodology Data cleaning/normalisation & processing Classification System WoS: 249 subject categories (incl. SSCI, AHCI) Scopus: 334 fields in ASJC (in 28 categories) Per Journal; overlapping Science Metrix (SM): 176 sub-fields (in 22 fields) Per Journal; disjunct SM provides classification of about journals Journals appear in different spellings (match or add? Complex SQL-manipulations) Classifications Editor & flexible handling of classification systems ~ 27 k Journals to classify, ~ 6000 of which were manually classified so far, only about 2000 organisations strings were edited 13

14 General data Records retrieved: 1,028,245 (all DR countries overall output except Germany s) The average DR publication (excl. Germany only) involves: 7.2 authors, affiliated in institutions in 1.8 countries, and is cited on average 6.7 times. Of the slightly over 1m records that were retrieved, there are 388,321 (citable) Danube-region co-publications with at least two of the Danube region countries involved. Each of these publications is cited an average of 12.1 times and involves 13.6 authors from 3 countries.

15 Data sources Records have been retrieved from both Scopus (Elsevier) and Web of Science (Thomson Reuters) WoS only Scopus only both

16 Presentation of the study methodology Analysis results tree by year by country by subfield by subfield by country data-set by country by year by subfield by subfield by country by subfield by year by country by sountry by subfield 16

17 Presentation of the study methodology Limitations Data cleaning not fully operated yet (status: interim results after unification) Classification systems comparability of e.g. field-normalised citation rates analysis on deeper level, e.g. city, affiliation, and especially author level outrageously high additional cleaning effort necessary references and detailed network analyses on big scale 17

18 Presentation of the study methodology Comments? Questions? Ideas? discussion 18

19 Language record = a publication (article, letter, etc.) indexed in Scopus or Web of Science co-publication = internationally co-authored indexed record impact = impact in the research community; citation counts are used as a proxy Danube region as defined in Danube-INCO-NET; for this presentation, taking into account the geographical distribution of the participants, the following countries have been in the focus: AT, BG, CZ, DE, BiH, HR, HU, MD, ME, RO, RS, SK, SI, UA STOP ME IF NECESSARY

20 General data Records retrieved: 1,028,245 (all DR countries overall output except Germany s) The average DR publication (excl. Germany only) involves: 7.2 authors, affiliated in institutions in 1.8 countries, and is cited on average 6.7 times. Of the slightly over 1m records that were retrieved, there are 388,321 (citable) Danube-region co-publications with at least two of the Danube region countries involved. Each of these publications is cited an average of 12.1 times and involves 13.6 authors from 3 countries.

21 Data sources Records have been retrieved from both Scopus (Elsevier) and Web of Science (Thomson Reuters) WoS only Scopus only both

22 Question What is the publication output in the Danube region and how is it developing over time?

23 Findings Output patterns suggest three country groupings: Austria, Czech Republic, Romania, Hungary and Ukraine (between 8,000 and 25,000 publications per year) Serbia, Croatia, Slovenia, Slovakia, Bulgaria (1,000-8,000) Bosnia and Herzegovina, FYROM, Albania, Moldova, Montenegro (100-1,000) Growth at all levels Particularly strong growth pattern in Romania and Serbia Relatively limited output growth in Hungary and Ukraine

24 Data Danube region publication output, ALB AUT BGR BIH CZE HRV HUN MDA MKD MNE ROU SRB SVK SVN UKR

25 Data 9000 Danube region publication output (excl AT, CZ, HU, RO, UK), ALB BGR BIH HRV MDA MKD MNE SRB SVK SVN

26 Data 1200 Danube region publication output (excl AT, CZ, HU, RO, UK), ALB BIH MDA MKD MNE

27 Findings In general, the shares of internationally co-authored publications are between 40% and 50% The general rule of a higher internationalisation of smaller research communities is confirmed However, there are some outliers: particularly internationalised, in addition to MD and ME (57-58%): Austria (50%) less internationalised: HR (26%), RO (26%), CZ (33%), UA (34%)

28 Question To what extent is it internationalised WITHIN the Danube region?

29 Findings In most DR countries, the share of inner-dr co-publications in all international co-publications amounts to 40-50%. These numbers are considerably higher in BiH and ME and slightly higher for HR. Germany, but also Romania and Ukraine have a lower share of DR co-publications in co-publications. We have to take into account here, that Germany and Ukraine are covered entirely at the moment! Removing the non-dr regions in these countries will reduce these figures.

30 Data country % int. copublications % of copub. in DR AT BG BiH CZ DE HR HU MD ME RO RS SK SI UA data for Germany from Scimago.com

31 Question What are the strongest co-publication linkages in the Danube region?

32 Findings It depends! Strongest according to absolute number of co-authored publications: Country A Country B % of copub. in DR AT DE CZ DE HU DE UA DE 6812 RO DE 6311 CZ SK 6220 BG DE 4846 SK DE 4491 AT CZ 4318 AT HU 3644 SI DE 3603 HR DE 3507

33 Findings Strongest according to shares of one involved country s overall output: Country A Country B copubs as % of A s as % of B s output output MNE SRB ,26% 1,17% AUT DEU ,43% 2,88% SRB BIH 842 1,67% 13,01% BIH HRV ,89% 1,40% CZE SVK ,57% 12,05% DEU BGR ,33% 12,03% DEU MDA 480 0,03% 11,50% HUN DEU ,59% 0,76% SVK DEU ,70% 0,31% CZE DEU ,99% 0,96% DEU UKR ,47% 7,28% MKD SRB 465 7,16% 0,92%

34 Findings Strongest when considering both countries overall output (Salton measures): Country A Country B copubs S as % of A s as % of B s output output AUT DEU , ,43% 2,88% CZE SVK , ,57% 12,05% MNE SRB 589 0, ,26% 1,17% SRB BIH 842 0, ,67% 13,01% BIH HRV 834 0, ,89% 1,40% SVN HRV , ,92% 3,36% HUN DEU , ,59% 0,76% CZE DEU , ,99% 0,96% MKD SRB 465 0, ,16% 0,92% AUT HUN , ,70% 3,49% SRB SVN , ,40% 2,38% AUT CZE , ,01% 2,47%

35 Question What are the most important DR partner countries for each DR country?

36 Findings Little surprisingly, DE, AT, CZ and HU are the most important partners in most cases Other evidence (snapshots): For BiH: RS, HR and SI more important than these four Strong CZ-SK link For HR: SI more important than AT, CZ and HU For MD: RO second most important after DE, UA third For ME: RS most important partner; HR third after DE; RO fourth For RO: HU second-most important; side note: FR most important partner in ERA!

37 Data Country Country B Copubs A-B Average cit. counts BGR DEU ,66 BGR CZE ,99 BGR HUN ,5 BGR AUT ,74 BGR ROU ,38 BGR SRB ,53 BGR UKR ,95 BGR HRV ,63 BGR SVK ,28 BGR MKD 263 6,81 BGR SVN ,01 BGR MNE 77 16,44 BGR BIH 49 9,51 BGR MDA 23 3,65

38 Question How did the DR co-publications develop over time?

39 Findings Growth across all copub-links A few links of particularly strong growth (identified through benchmarking the growth in DR copublications of a country with its general growth pattern): AT: CZ, RO, RS, BG BG: CZ, RS, AT, UA, HR BiH: RS (since 2010 higher outputs with RS than with HR), HR CZ: RO, HR, HU, SK, AT HR: RS, RO, HU, CZ, BG HU: RO, RS, AT, CZ MD: RO ME: RS, HR RO: HU, CZ, AT, RS, UA RS: HR, SI, HU, AT, CZ, BiH, RO SI: RS, RO, HU, SI SK: RS, RO, SI, CZ UA: RO, RS, BG

40 Data RO ex DE RO-RS RO-SK RO-SI RO-UA ME-RO MD-RO HU-RO HR-RO CZ-RO BiH-RO BG-RO AT-RO

41 Question What is the thematic collaboration portfolio in the Danube Region?

42 Findings Comparing the thematic distribution of DR copublications vs output allows the following observ.: Stronger (in relative terms!) in copubs than in pubs: physics & astronomy Chemistry biomedical research Biology earth & environmental sciences Weaker in copubs than in pubs: Engineering ICTs

43 100% 90% General Arts, Humanities & Social Sciences Visual & Performing Arts Built Environment & Design Philosophy & Theology 80% Communication & Textual Studies Public Health & Health Services 70% 60% 50% 40% 30% Psychology & Cognitive Sciences General Science & Technology Historical Studies Social Sciences Economics & Business Earth & Environmental Sciences Agriculture, Fisheries & Forestry Mathematics & Statistics Biology Biomedical Research Chemistry 20% 10% Enabling & Strategic Technologies Information & Communication Technologies Engineering Physics & Astronomy 0% records DR overall DR copubs (citables) Clinical Medicine

44 Question How does the thematic portfolio of DR countries relate to the DR thematic portfolio?

45 Findings Two levels are relevant for benchmarking here: DR copub output compared to country copub output DR copub output of a country compared to the DR copub output in general Combination and qualitative analysis of these A set of countries (AT, BiH, HU) is relatively more specialised in the area of clinical medicine in their DR collaboration Others (BG, CZ, MD, RO, SK, SI, UA) have stronger DR links in the area of physics & astronomy [likely big science collaborations]

46 Findings Some additional punctual specialisations like: AT in biomedical research, earth & env sciences [national strength in ICT is not reflected] BG in enabling & strateg technologies [not as strong as in BG copubs], chemistry BiH in engineering, biology and agric./food/fisheries, social sciences and historic. studies CZ in chemistry, biomedical research (although less than in its copubs in genera], biology, earth & env sciences HR in chemistry, biomedical research, historical studies HU in biomedical research (also less than in its copubs in general), earth & env sciences MD in enabling & strat tech, especially chemistry ME in chemistry, engineering (less than in general copubs), biology RO in chemistry RS in chemistry; general copub strength in math is not reflected in RS-DR SK in chemistry and biomedical research (biomed to a lesser degree than in nat copubs) SI in enabling & strat tech, chemistry and agric/food/fisheries UA [apart from dominant physics] in chemistry, enabling & strat tech

47 Data - RS Science Metrix fields (most important in DR) RS overall (citables) share RS copubs share m ccit RS copub RS-DR share Clinical Medicine ,0% ,8% 10, ,7% Physics & Astronomy ,6% ,7% 12, ,2% Engineering ,6% 966 6,2% 5, ,0% ICTs ,8% 960 6,2% 4, ,9% Enabling & Strat. Tech ,0% ,7% 6, ,5% Chemistry ,4% ,4% 8, ,9% Biomedical Research ,3% 712 4,6% 9, ,4% Biology ,2% 875 5,7% 6, ,6% Mathematics & Statistics ,3% ,6% 7, ,5% Agric., Fish. & Forestry ,6% 505 3,3% 5, ,1% Earth & Env. Sciences 937 1,9% 323 2,1% 7, ,4% Economics & Business 680 1,3% 170 1,1% 2, ,1% Social Sciences ,6% 238 1,5% 2, ,9% Historical Studies 313 0,6% 122 0,8% 10, ,0% General S&T 478 0,9% 149 1,0% 8, ,9% Psychol. & Cognit. Sc ,9% 110 0,7% 9, ,5%

48 Question How do the thematic specialisations develop over time (in relation to the overall development of output)?

49 Findings This is a question that might be particularly relevant in the case of countries where output growth is particularly strong (like Romania or Serbia) to clarify whether the growth is sustained over all topics. For the examples mentioned, findings include: RO: strong growth can be observed in the areas of biomedical research, earth & env sciences, enabling & strat tech, economics & business, engineering, historical studies, ICTs, psychology, social sciences; less in agric./food/fish., chem. or public health RS: strong growth in biology, biomed research, clinical medicine, economics & business, enabling & strat tech, engineering, social sciences physics is not the main driver behind the output growth! Similar data can be extracted for all countries.

50 Question What is the impact of Danube region collaboration?

51 Findings The question of the impact of the research output in the Danube region becomes relevant on several levels: Impact of output of countries Impact of collaborative output of DR partners (by pair) Impact of DR countries in specific topics Impact of DR partners output in specific topics Not all data can be presented here, but a few snapshots General country-level: As stated above, each citable Danube region record covered in this study is cited on average 6.7 times. The countries with the highest average citation counts in their general output are 1 : Hungary: 9.16 Austria: 7.97 Slovenia: Germany is not considered here as not all Germany records have been retrieved

52 Findings The DR collaboration country pairs with the highest impact and a co-published output >250 ( ): RS-UA (27.8; 477 records) UK-HU (26.8; 859 records) CZ-HR (24.7; 1,175 records) HR-UA (24.0; 550 records) BG-HR (23.6; 603 records) CZ-HU (23.5; 2,484 records) DE-HR (23.1; 3,507 records) Given that the thematic area of collaboration influences the average impact count, a separate consideration by topic is recommendable (field normalisation!)

53 Findings Compared to the av. times cited counts in DR collab. output (388,321 citable records) by them. area, country strengths noted 1 : AT: physics & astronomy BG: clinical medicine; physics & astronomy BiH: physics & astronomy CZ: clinical med.; physics & astronomy; ICTs; enabling & strat tech; biology; psychology & cognitive sciences; public health & health services; historical studies; social sciences HR: physics & astronomy; biomedical research; general S&T HU: physics & astronomy; clinical medicine; biomed. research; earth & env sciences; economics & business; social sciences; historical studies; psych. & cognit.; public health ME: physics & astronomy RO: clinical medicine; physics & astronomy; public health RS: physics & astronomy SK: ICTs; mathematics & statistics; economics & business; public health SI: physics & astronomy; ICTs; general S&T; public health UA: clinical medicine 1 we only consider thematic areas with a minimum of 100 records

54 Question Questions?

55 Group discussions Country-data is available during the coffee break (country profiles in print and hand-outs of ppt) Suggestions for group discussion: General comments to the data presented? Country-data: Particularities? Things you found noteworthy? Questions for clarification that might arise? Your country-expertise: Trends/developments in the country s research community (including policy or funding) that we should take into account?

56 Next steps Among the next methodological and analytical steps are: Sub-national level of DE and UA (länder and oblasts) Continuation of benchmarking exercise Analysis of cooperation within ERA and beyond Organisation-level data cleaning and analysis Following up particularities in national profiles when digging deeper into the data Separate consideration of the big science areas (like physics & astronomy) Advanced visualisations

57 A. Degelsegger / D. Lampert Centre for Social Innovation degelsegger@zsi.at lampert@zsi.at Thank you for your attention!

58