Negative binomial regression models predicting the ex-post number of total articles and their citations for awarded applicants (vs. rejected applicants as reference group).

Note. The link between dependent variables (“ex-post number of total articles” and “ex-post total citation counts”) and independent variables is presented as regression coefficients (standard error). Model 1 – an independent variable “applicants” only (where 0 stands for rejected and 1 for awarded applicants); Model 2 – Model 1 + “ex-ante number of total articles” and “ex-ante total citation counts”: Model 3 – Model 2 + all factors from Table 1 and covariate “Rank of current institution”. The category “immunology” of the variable “specialty” was not taken into account in the regression analysis due to the absence of awarded applicants with this specialty. Five missing values of the variable “rank of current institution” are replaced with the maximum values of the attribute [rank = 1486] in the sample. Thus, these institutions were given low positions in the eLIBRARY.RU institute ratings (the argument – the position in the ratings is not assessed for institutions whose employees publish few articles). No statistically significant link between the independent variable “applicants” and the dependent variables “ex-post number of total articles” and “ex-post total citation counts” was detected in any of the regression models (in all of the cases p>0.05).