Empirical Probability Distribution of Journal Impact Factor and Over-the-Samples Stability in its Estimated Parameters

The data on JIFs provided by Thomson Scientific can only be considered as a sample since they do not cover the entire universe of those documents that cite an intellectual output (paper, article, etc) or are cited by others. Then, questions arise if the empirical distribution (best fit to the JIF data for any particular year) really represents the true or universal distribution, are its estimated parameters stable over the samples and do they have some scientific interpretation? It may be noted that if the estimated parameters do not exhibit stability over the samples (while the sample size is large enough), they cannot be scientifically meaningful, since science is necessarily related with a considerable degree of regularity and predictability. Stability of parameters is also a precondition to other statistical properties such as consistency. If the estimated parameters lack in stability and scientific meaning, then the empirical distribution, howsoever fit to data, has little significance. This study finds that although Burr-4p, Dagum-4p and Johnson SU distributions fit extremely well to the sub-samples, the parameters of the first two distributions do not have stability over the subsamples. The Johnson SU parameters have this property