This memorandum stipulates three sampling standards to be
followed in the conduct of the BRFSS survey.
First, the sample used should be justifiable as a
probability sample from a sampling frame consisting of all households with
telephones. One implication of this standard is that purchased samples
must include telephone numbers selected from all working banks of 100
telephone numbers that may contain household numbers, not just those
containing 1, 2, 3, or more listed household numbers. A further
implication of this standard is that the sampling frame for purchased
samples must include telephone numbers from the entire state and thus must
include counties and other geographical areas with a small number of
residences. It is permissible, however, to sample telephone numbers from
different working banks or different parts of a state at different rates,
provided that the sampling ratio is reported to BSB.
Second, states using the Mitofsky-Waksberg sampling method
should use the same procedures and rules of replacement in conducting
pre-screening to identify eligible clusters, or primary sampling units, as
they do in selecting households within clusters. The validity of the
Mitofsky-Waksberg sampling method depends upon using the exact same
sampling procedures and rules of replacement at both the cluster and
household selection stages. A satisfactory procedure that several states
have found convenient is to conduct pre-screening for the following
month's sample at the same time that they conduct the household interviews
for the current month.
Third, it is recommended that states using the
Mitofsky-Waksberg sampling method not identify themselves or the purpose
of the survey during pre-screening, unless they interview the respondent
immediately. Such self-identification biases the survey by giving
potential respondents a month to think about the survey topic and their
willingness to cooperate when they are contacted the following month.
