554 – SPAAC Poster Competition
Break Detection Methods Applied for International Industrial Property Time-Series Data
BeomYong Kim
Seoul National University
The US changed its patent policy toward pro-patentism in 1980s. Japan discarded one of dual IP(Industrial Property) infringement lawsuit systems in 1960. Korea adopted material(composition of matter) patent in 1987. These 3 break cases can be detected by iteratively moving QLR(Quandt Likelihood Ratio) tests thru using patent or the like time-series data. 1st, using stata command "varsoc", we can select the lag-orders for the level data themselves or the natural log data of the level data. We may have to use the difference data from the above data. The selection standards are FPE, AIC, HQIC, SBIC. 2nd, using the command "cusum6", we can select the appropriate & final model from the above candidate models. Selection standards are the cumulative sums(CUSUM) of the recursive residuals and their squares from the above models' regressions. 3rd, in applying the general regression to the time series data we tend to exaggerate the both ends and so should adopt the centered 70% range. The dummy variable "di" indicates the point of the break. The multiplications of the dummy variable "di" and independent variables of the above final model are required for the above QLR test coding contents.