Logistic Regression Modeling of Diminishing Manufacturing Sources for Integrated Circuits

摘要

The problem of Diminishing Manufacturing Sources and Material Shortages (DMSMS) afflicts the Air Force and other armed services in a way unforeseen in the past. The costs of DMSMS manifest themselves both in terms of monetary expense and diminished mission readiness. The current environment of limited defense spending and high operations tempo exacerbates this problem. This thesis draws on available data from the electronics integrated circuit industry to attempt to assess whether statistical modeling offers a viable method for predicting the presence of DMSMS. The research identified logistic regression as a powerful tool for analysis of DMSMS and further developed twenty models attempting to identify the 'best' way to model and predict DMSMS using logistic regression. The results indicated that DMSMS does not seem related to technical characteristics. However, design age proved to be statistically significant in modeling DMSMS presence. Another finding of this research indicates that integrated circuits designed for military use have extended longevity compared to commercial integrated circuits. The overall conclusion of this effort was that DMSMS can be predicted using the conditional probability output of logistic regression. Current databases do not provide easy access to the vital predictor, design age. Recommendations include further validation of the model and the inclusion of design age in electronics integrated circuit databases.