Robust Regression Methods For Massively Decayed Intelligence Data

摘要: ROBUST REGRESSION METHODS FOR MASSIVELY DECAYED INTELLIGENCE DATA by AKIVA JOACHIM LORENZ May 2014 Advisor: Dr. Barry Markman Major: Evaluation and Research Degree: Doctor of Philosophy Homeland Security, sponsored governmental initiatives, has become a vibrant academic research field. However, most efforts were placed with the recognition threats (e.g. theory) response options. Less effort was in analysis collected data through statistical modeling. In field that collects more than 20 terabyte information per minute though diverse overt covert means indexes it for future research, understanding how different models behave when comes to massively decayed is vital importance. Using Monte Carlo methods, three regression techniques (ordinary least squares, least-trimmed, maximum likelihood) tested against decay presumed be found homeland security studies order test whether these will preserve Type I error rate t-test standardized beta. The results simulations (sample size n=30,90,120,240,480 100,000 iterations) showed trimmed squares method should avoided under any circumstance due lack defined standard error, while likelihood technique smaller sample sizes inflated errors. Interestingly, although known ordinary