Rivero, CarlosValdés Sánchez, Teófilo2023-06-202023-06-202004-090303-689810.1111/j.1467-9469.2004.01_108.xhttps://hdl.handle.net/20.500.14352/50493We present in this paper iterative estimation procedures, using conditional expectations, to fit linear models when the distributions of the errors are general and the dependent data stem from a finite number of sources, either grouped or non-grouped with different classification criteria. We propose an initial procedure that is inspired by the expectation-maximization (EM) algorithm, although it does not agree with it. The proposed procedure avoids the nested iteration, which implicitly appears in the initial procedure and also in the EM algorithm. The stochastic asymptotic properties of the corresponding estimators are analysed.engjournal articlehttp://www.jstor.org/stable/10.2307/4616843http://www.jstor.org/restricted access519.2Censored-datamaximum-likelihoodem algorithmregressionasymptotic distributionsconsistencyexpectation-based imputationgrouped dataiterative estimationlinear modelsnested iterationEstadística matemática (Matemáticas)1209 Estadística