Post-Graduate
Applied and Computational Statistics Group
Project Description
NIST statisticians recently devised an innovative regression slope estimator based on an artificial neural network (ANN). This project will extend this ANN-based method of estimation into different domains to better understand its potential. These domains will include joint estimation of intercept and slope in linear regression and in logistic regression, and will consider both uniformly and non-uniformly spaced regression design points. Different ANN architectures will be explored to test the sensitivity of ANN estimation to architecture choice. And, architectures will be sought for which the ANN slope estimators bias is identically zero by construction.