Patents from 1987, cited through to 2006
It is possible that received citation profiles change over time with results depending on the time interval chosen. We pursue this idea here by using a dataset defined by patents granted in 1987. While this reduces the number of years over which received citation distributions are defined, there is an additional reason for performing this analysis. We were able to include patent heterogeneity and self-citations as two more supplementary variables became available. A total of 12,291 patents granted in 1987 were cited at least 20 times to the end of 2006. The time interval for citations is 20 years. The methodology used is exactly the same as in subsection 5.7.1. After establishing 40 leader profiles, these were clustered into four clusters. The resulting dendrogram and the corresponding four leader profiles are in Figure 5.24. The first cluster, on the left of Figure 5.24, features patents in which interest increased continuously until 2004 before dropping. Interest still remained high in 2006. Patents in the second cluster had a sharp increase in interest until 1989 when interest in them reached a peak. Thereafter, there was a decline to a modest level of interest in 2006. Interest in the patents in the third cluster rose steadily through 1998. This was followed by a sharper decline in interest than for the patents in the second cluster. Interest in the patents in the fourth cluster also rose sharply, reaching a peak shortly after 1989. This interest was maintained at a (slightly lower) high level through 1995. Interest in these patents then dropped to near zero. In the 2000s, this interest dropped to zero.
Figure 5.25 shows the actual temporal profiles of four patents with one from each 1987 patent cluster as in the same order of the clusters shown as in Figure 5.24. The first patent, number 4,663,769, was issued to Motorola Inc. for dealing with asynchronous communication with a clock acquisition indicator circuit for non-return-to-zero data. Belonging to the method and computers and communication categories, its spike in interest occurred at the end of the time period considered. The second patent, number 4,692,352, was awarded to a Japanese corporation, Matsushita Electrical Industrial, for a method and apparatus for drawing a thick film circuit. Belonging to the method and the electrical and electronic categories, it had the reverse profile of an early peak in citation followed by a decline in citations.
The third patent in Figure 5.25, number 4,673,695, was granted to the USA Government for the invention of low density, microporous polymer foams created by a solution of a polymer and a solvent through a rapid cooling process. This patent belongs to the composition of matter and chemical categories with a profile having a steady growth in citations to a peak followed by a sharp decline. The final profile in this figure is for patent number 4,650,479 awarded the MMM Company for a sorbet sheet product useful for disposable diapers, incontinence devices, and sanitary napkins. Belonging to the composition of matter
Figure 5.24 Dendrogram for hierarchical clustering on the 40 leaders with <53 as the error measure, 1987-2006.
The four summary leader patterns are shown below the corresponding red boxes of the dendrogram
Figure 5.25 Example profiles of patents in each identified 1987 patent cluster.
Table 5.5 p-values for Kolmogorov-Smirnov statistical tests for the four final 1987 clusters.
Statistically significant results are marked by * (with a = 0.005).
and drugs and medical (miscellaneous) categories, it had an immediate impact in terms of citations followed by a decline to zero.
Utilizing supplementary variables for 1987 patents
The results of considering the supplementary variables for these 1987 data are shown in Table 5.5. As for the 1976 data, all differences between the clusters for the number of citations and the overall distribution are significant. The plots of citation numbers for the four clusters are shown on the left of Figure 5.26. Regarding generality (being cited by patents in different technological areas), two clusters - the first and fourth clusters - are significantly different from the overall citation level. From the right panel of Figure 5.26, the first cluster has higher generality while the fourth cluster has lower generality. In this regard, the fourth cluster is like cluster 3 of the 1976 data. Another difference between the results for the 1976 and 1987 patents is a contrast between cluster 1 from 1976 and the second cluster for 1987. Both clusters had early peaks in their interest value. However, cluster 1 in the 1976 data had a higher citation frequency while the second cluster of the 1987 had a lower such frequency. Finally, the supplementary variables heterogeneity and self-citation levels by patent authors to their earlier patents do not differ across the four clusters.
-  por greater visual clarity, the empirical cumulative distribution functions are plotted for only the first and fourth clusters. The distribution of generality measures for the other two clusters do not differ from the overall distribution.