A general formulation of bradford's distribution: The graph‐oriented approach

From the detailed analysis of eight previously published mathematical models, a general formulation of Bradford's distribution can be deduced as follows: y = a log( x + c ) + b , where y is the ratio of the cumulative frequency of articles to the total number of articles and x is the ratio of the rank of journal to the total number of journals. The parameters a, b , and c are the slope, the intercept, and the shift in a straight line to log rank, respectively. Each of the eight models is a special case of the general formulation and is one of five types of formulation. In order to estimate three unknown parameters, a statistical method using root‐weighted square error is proposed. A comparative experiment using 11 databases suggests that the fifth type of formulation with three unknown parameters is the best fit to the observed data. A further experiment shows that the deletion of the droop data leads to a more accurate value of parameters and less error.