Comment on `Open is not forever: a study of vanished open access journals'
aa r X i v : . [ c s . D L ] N ov Comment on ‘Open is not forever: a study of vanished openaccess journals’
Gerta R¨uckerInstitute of Medical Biometry and Statistics, Medical Center – University of FreiburgStefan-Meier-Strasse 26D-79115 Freiburg, Germanye-mail: [email protected] is a comment to an article by Laakso, Matthias and Jahn (arXiv:2008.11933).With interest I read an article published by Laakso et al. (1) that was already commentedon by Shelomi (2). Laakso et al. reported a study in which they found 176 Open Access(OA) journals that had vanished from the web between 2000 and 2019. As Shelominoted, it was unfortunate that Laakso et al. did not distinguish between predatory andreputable journals, and I agree. In my present comment, however, I want to point toanother weakness of the article by Laakso et al., which is the statistical analysis. Theyconsider their sample of 176 vanished journals and present some descriptive statisticson their characteristics, such as the time lag between the last publication and vanishing(their Figure 3 and Table 3), the distribution of academic disciplines (their Figure 4), theiraffiliation (whether academic or not), and the geographic distribution (Figure 5 and Table4). Of course, this is legitimate as a description of a selected sample. However, the authors,ignoring that their sample was selected conditionally on a journal having vanished fromthe web, misinterpret their findings. Their statement that ‘our study provides valuableinsight into the types of OA journals that are especially at risk of vanishing’ (page 22)is not justified by evidence, simply because a risk can only be quantified on a populationbasis. Otherwise, any analysis can only be descriptive.Laakso et al. found that ‘North America and South Asia represent a disproportionatelylarger share of vanished than active OA journals’ and interpreted this as ‘journals published1n North America . . . belong to the high risk group’. For the geographic distribution thisis justified, as they refer to a list of active OA journals as a control group (Table 4).However, they use the term ‘risk’ also for other factors, writing for example: ‘Moreover,our study provides valuable insight into the types of OA journals that are especially at riskof vanishing’, then referring to journals affiliated with academic institutions or scholarlysocieties, without providing the numbers of active OA journals for comparison. Thismight provoke a well known and common error: the inversion fallacy, that is, confusingthe denominators. To find out whether scholarly-based journals are at higher risk ofvanishing than other journals, it is not correct to compare the proportions of journals byaffiliation within vanished journals . It would be necessary to compare vanished and activejournals and to adjust for covariates in a multivariable analysis. The limitation of thisstudy as a case series should have been acknowledged by the authors.