Temporal changes in the parameters of statistical distribution of journal impact factor

Statistical distribution of Journal Impact Factor (JIF) is characteristically asymmetric and non-mesokurtic. Even the distribution of log10(JIF) exhibits conspicuous skewness and non-mesokurticity. In this paper we estimate the parameters of Johnson SU distribution fitting to the log10(JIF) data for 8 years, 2001 through 2008, and study the temporal variations in those estimated parameters. We also study ‘over-the-samples stability’ in the estimated parameters for each year by the method of re-sampling close to bootstrapping. It has been found that log10(JIF) is Pearson-IV distributed. Johnson SU distribution fits very well to the data and yields parameters stable over the samples. We conclude that Johnson SU distribution is the best choice to fit to the log10(JIF) data. We have also found that over the years the log10(JIF) distribution is becoming more skewed and leptokurtic, possibly suggesting the Mathew effect in operation, which means that more cited journals are cited ever more over time.