(Zero-truncated) negative binomial regression models predicting (1) number of papers published subsequent to application, (2) citations for papers published prior to application and (3) citations for papers published subsequent to application (applicants for the LTF programme).
Note. ML-point estimates (the results of the z-test in parentheses).
*p<0.05, **p<0.01, ***p<0.001.
1truncated sample.
There is one paper in the sample for model 3 with an exorbitant number of co-authors (n = 2,458) (see Table 1). Omitting this paper from the regression analysis did not alter the statistically significant coefficient for the variable “Decision” that is presented in the table.
Interpretation example for the parameter estimates in the table: In model 2 the number of pages of a publication has a statistically significant effect on receiving citations with a parameter estimate of 0.04. This means that for an additional page, the odds of receiving citations increase by a factor of 1.04 ( = exp(0.04)), holding all other variables in model 2 constant.