Some large trial properties of minimum likelihood allocation (Q1057613)

The paper presents a formal description of minimum likelihood allocation method and proves two of its ''long run'' properties: ''consistency'' and ''efficiency''. ''Consistency'' is connected with the fact that the average likelihood ratio which measures design imbalance tends to zero and is proved using game theory methods. ''Efficiency'' concerns the asymptotic variance of parameter estimates. The model under consideration is the linear exponential model and can be found in the author's book ''Linear statistical analysis of discrete data''. Computer support for MLA can be obtained from the author.