[The intercensal comparability problems in the retrospective infant mortality statistics].

In infant mortality estimates, using the Coale-Trussell procedure with information from 2 successive population censuses of a country, reveals differences in the levels for the overlapping years of periods. This paper deal with this problem using the 1974 and 1982 censuses of Ecuador. The contribution of his analysis consists in demonstrating that infant mortality in a given year--estimated from a specific age group--is a weighted average between zero mortality--women declaring no deaths among their children and mortality with a value different from zero--women declaring some child dead. The weighing values are the age structure of women according to the condition of having dead children and corrected by the mean parity of the corresponding age group. Looking at the structure--women with or without dead children--of both censuses, a strong increase in the proportion of women without dead children occurred in 1982 as compared with the proportion in 1974. In order to eliminate this lack of comparability a reweighing was introduced in the 1982 information. New estimates were obtained, and these are much closer to the estimates obtained from the 1974 census.