Tests for underlying Markovian structure from panel data with partially aggregated states
部分状态聚合面板数据中潜在马尔可夫结构的检验
SUMMARY For dichotomous longitudinal panel data assumed to have been generated by a Markov chain but with only one of its states identifiable, the likelihood ratio and related tests for underlying Markovian structure are studied. It is shown that the likelihood ratio may be written as a product of conditionally independent likelihood ratios for testing independence in certain contingency tables, given margins. The tests are illustrated on data from malaria and…