Statistical studies of patents literature
This study has been undertaken to determine what pseudo-proprietary information and patenting activity statistics could be derived from an online patents database. To achieve this, a thorough investigation uas made of patenting in the field of an important group of beta-lactam antibiotics, the. Cephalosporins. Patents data was retrieved from the World Patents Index online files of Derwent Publications Limited, and the bibliographic details of each patent application retrieved analysed according to numbers of patents per patentee, priority and publication dates, types of patents, etc. A review of technological advances in this subject was conducted, demonstrating the value of patents literature for such purposes. The relationship between sales volumes and patenting activity for Cephalosporins patentees has been investigated and found to show a. significant correlation between these parameters. As an extension, the USA patenting and sales activity for the leading USA Industrial Corporations (the 1981 Fortune 500) was studied; overall a high correlation was exhibited, but there were notable differences. between different industries. A number of bibliometric studies have been undertaken with a variety of patents data. for a number of techhologies. These studies include the application of Bradford-Zipf plots, other productivity studies and Vector Analysis to patents. Whilst previous studies on journal literature have investigated the applicability of frequency distributions as measures of author productivity, this study has for the first time applied Lotka's Law, Price's Pareto-type Distribution, Simon-Yule Distribution, Shockley's Lognormal Distribution, Borel-Tanner Distribution, Williams Geometric Series, Fisher's Logarithmic Series and the Negative Binomial Distribution to patents data. Theoretical distributions were ascertained using a series of microcomputer programs written in BASIC programming' language. The results indicate that of the distributions investigated, the Negative Binomial most closely fits the observed data when goodness-offit is measured by the Kolmogorov-Smirnov Test.