Sample characteristics: Brazil

Also included: population in transit (those on ships, planes, trains and buses) and people in collective dwellings. Brazilians residing abroad, even those who were performing official governmental functions, were not enumerated.

The indigenous population was only partially enumerated - only those living in villages in contact with civilization.

Each collective dwelling had a separate list form - CD 3 - (Lista de Domicílio Coletivo) to select the families or individuals for the sample.

A 25% sample was taken of individual dwellings and of families or individuals in "grupo conviventes".

**NOTE: Several states totaling approximately one sixth of the national population are missing from the long-form microdata, and are thus not included in the 1960 sample data. These states are Rondônia, Acre, Amazonas, Roraima, Pará, Amapá, Maranhão, Piauí, Espírito Santo, Guanabara, and Santa Catarina.

The IPUMS sample assumes a weight appropriate to the theoretical 25% sample density for the long-forms.

Also included: Brazilians on diplomatic and military missions abroad; crew and passengers of Brazilian ships that were not in domestic ports.

Not included: members of foreign diplomatic or military missions in Brazil who lived in embassies on Brazilian soil; crew and passengers of foreign ships in transit in domestic ports; indigenous population living in tribes without direct and permanent contact with civilization.

Each collective dwelling had a separate list form - CD 1.03 - (Lista de Domicílio Coletivo) to select the families or individuals for the sample.

A 25% sample was taken of individual dwellings and of families or individuals in "grupo conviventes".

When necessary, some control groups were added together (i.e., when the expansion factor was greater than 16 or the control group universe was less than 100).

Each person received an integer weight such that final mean of the sample attained the true fraction of the universe. For example, if 140 people were in the universe, 42 people would be included in the sample, which gives an overall sampling ratio of 3.333. 28 persons were randomly selected to receive a weight of 3 and 14 received a weight of 4 such that a final mean of 3.333 was attained.

For estimates of family characteristics, the weight used is that attributed to the head of the family; for estimates of household characteristics, the weight used is that attributed to the head of dwelling.

Also included: Brazilians (and their families) on diplomatic and military missions abroad; crew and passengers of Brazilian ships that were not in domestic ports; indigenous population living in villages, indigenous posts (postos indigenas), and religious missions, or in other areas.

Not included: indigenous population living in tribes without contact with the outside, those preserving their primitive ways of life); foreigners on foreign ships in transit in domestic ports or domestic airports.

Each collective dwelling had a separate list form - CD 1.03 - (Lista de Domicílio Coletivo) to select the families or individuals for the sample.

A 25% sample was taken of individual dwellings and of families or individuals in "grupo conviventes".

Matrix marginals were used to calculated the ratio estimators (rather than those for each cell).

Each person received an integer weight such that the final mean of the sample attained the true fraction of the universe, as in 1970.

Also included: indigenous population living in villages, indigenous FUNAI posts (postos indigenas), and religious missions, or in other areas.

Not included: indigenous population living in tribes without contact with the outside (those preserving their primitive ways of life).

The long form was applied to a 10% sample of the population in municipalities with estimated populations greater than 15,000 and 20% in the remaining municipalities (Silva e Bianchini 1990).

There were two distinct sample fractions: a 10% sample in municipalities with an estimated population of greater than 15,000 inhabitants and a 20% sample of the remaining municipalities.

Each dwelling in the sample has its own weight, so each person of the dwelling has the same weight as the dwelling's weight.

The weights are not necessarily integers and they must not be substituted with integers because one would violate the consistence of the restrictions employed for adjustment.

Also included: people who were absent (for reasons of vacation, education, work, or hospital stays, for less than 12 months) from their dwelling.

Embassies, consulates and Brazilian representatives out of the country are considered National Territory; however, they are not included in the Census (the majority of Brazilian functionaries do not live on the grounds of diplomatic missions).

The long form was applied to a 10% sample of the population in municipalities with estimated populations greater than 15,000 and 20% in the remaining municipalities (Silva e Bianchini 1990).

There were two distinct sample fractions: a 10% sample in municipalities with an estimated population of greater than 15,000 inhabitants and a 20% sample of the remaining municipalities.

This method is based on a multivariate procedure in twostages developed by Bankier (1990).

Each dwelling in the sample has its own weight, so each person of the dwelling has the same weight as the dwelling's weight.

The weights are not necessarily integers and they must not be substituted with integers because one would violate the consistencies of the restrictions employed for adjustment.

Note: In order to obtain consistency with the published Census tables, it is necessary that the estimates are calculated in each table cell and that the row and column totals and subtotals are obtained by summing the corresponding basic estimates, after they have been rounded. A consequence of this procedure is that the total for the same characteristics could vary slightly from one table to another because of summing rounded numbers from the cells. In the same way, the estimates for Brazil could differ from the value obtained through summing the corresponding published estimates for each state in the union.

Also included: people who were absent (for reasons of vacation, education, work, or hospital stays, for less than 12 months) from their dwelling.

Embassies, consulates and Brazilian representatives out of the country are considered National Territory; however, they are not included in the Census (the majority of Brazilian functionaries do not live on the grounds of diplomatic missions).

The long form was applied to a 5% sample of the population in municipalities with estimated populations greater than 500,000; 10% sample of the population in municipalities with estimated populations between 20,000 and 500,000; 20% sample of the population in municipalities with estimated populations between 8,000 and 20,000; 33% sample of the population in municipalities with estimated populations between 2,500 and 8,000; and 50% in the remaining municipalities.

There were five distinct sample fractions: a 5% sample of the population in municipalities with estimated populations greater than 500,000; 10% sample of the population in municipalities with estimated populations between 20,000 and 500,000; 20% sample of the population in municipalities with estimated populations between 8,000 and 20,000; 33% sample of the population in municipalities with estimated populations between 2,500 and 8,000; and 50% in the remaining municipalities.

The expansion of the sample used Generalized Minimum Least Squares with the imposition that the final weight could be no less than 1 and no greater than five times the initial weight of the effective level of a balancing area.

This method is based on a multivariate procedure in twostages developed by Bankier (1990).