On powers of likelihood ratio tests constructed from samples with random sizes

The conditions are presented for the powers of likelihood ratio tests constructed from samples with random sizes to have non-trivial limits. For these tests two interpretations are proposed for the concept of power. Particular examples are considered.