More just lately find out morea study have proven that the inter-rater agreement of peer reviewers for journal articles or blog posts is as lower as .23. Panel A guidance proposed that results should be informed by citation metrics but REF evaluators have been provided no standardised advice as to how it was to be connected to the score of three or four.To test whether or not the documented doubling of planet-foremost investigation high quality in the REF was supported by bibliometric top quality, we obtained particulars of all Panel A outputs submitted to the RAE 2008 and REF 2014, and when compared them to different bibliometric indices, alongside with international overall performance on the exact same indicators in that time frame.We acquired all the outputs submitted for Panel A and B for both RAE and REF from the general public website. The outputs submitted for REF have been categorised into journal articles or blog posts, chapters, guides and other kinds of research communications. Because citation metrics are most pertinent for journal articles or blog posts, we limited additional evaluation to these research outputs which shaped 97%/99% of the complete outputs in Panel A for RAE/REF. The RAE and REF modified the nomenclature of panels and subpanels and to ensure comparability, we transformed all the RAE panel data into REF panel terminology, as a result permitting a immediate comparison.The RAE/REF content articles had been matched from the Centre for Science and Technologies Reports, Leiden version of the Net of Science database which contains info on over forty two million posts from in excess of 18 thousand journals and tracks far more than 555 million citations. To examine like-with-like, the quotation to a distinct article was in contrast to all other articles or blog posts in the identical subject and from the same year of publication allowing the willpower of the throughout the world percentile of that post. Some content articles are in journals that relevant to far more than one particular class, in which scenario the all round standing of the write-up is the sum of its fractional standing in the fields to which it is assigned. We excluded self-citations in our examination as we wanted to measure externally obtained quotation impact, thereby excluding up-front any supply of distortion. Ultimately, to validate our use of bibliometrics we examined how generally the journal articles or blog posts in our RAE/REF set referred to other articles which ended up also in this dataset, a measure known as internal coverage, figures increased than 70% offer legitimate and secure steps of bibliometric indicators. Because the fields of Health-related and Life Sciences in this database are mostly coterminous with Panel A in RAE/REF, we attained bibliometric data relating to British isles vs. With this info available for all RAE and REF journal posts we questioned the subsequent queries: a) how numerous RAE/REF articles or blog posts have been offered for bibliometric examination b) how did the good quality of articles or blog posts submitted in the RAE vs. REF assess in terms of worldwide percentile standing c) how did top quality of all British isles Health care and Life Sciences content articles alter for the duration of that period d) has there been a reducing of the percentile thresholds at which papers are labelled three/4 and e) is this alter of threshold seen only in Panel A or also in Panel B .The REF Panel A acquired fifty,044 content articles, reflecting a slide in all round quantity from fifty six,639 acquired in RAE 2008. Of these, ninety six% of the RAE and ninety eight% of the REF posts ended up connected to the CWTS Net of Science databases and as a result amenable to bibliometric analyses.