Sample balancing is widely used in applied research to adjust a sample data to achieve better correspondence to Census statistics. The classic Deming-Stephan iterative proportional approach finds the weights of observations by fitting the cross-tables of sample counts to known margins. This work considers a bi-criteria objective for finding weights with maximum possible effective base size. This approach is presented as a ridge regression with the exponential nonlinear parameterization that produces nonnegative weights for sample balancing.
"Nonlinear Parameterization in Bi-Criteria Sample Balancing,"
Journal of Modern Applied Statistical Methods:
1, Article 19.
Available at: http://digitalcommons.wayne.edu/jmasm/vol9/iss1/19