-----BEGIN PGP SIGNED MESSAGE-----
Hash: SHA1

- - 16 -

Other general results of the survey, such as the average sobriety, the
average age, the percentage addicted to drugs, etc., have roughly this
kind of reliability, subject always to ambiguities associated with the way
in which sampling is done, questions are posed by the questionnaire, etc.
However, any results which are derived from a small fraction of the
population (for example, the sex of teenagers, or of a single area) have
limited validity.

Correction for Attendance Rates:   Obviously the sampling will pick up
disproportionately those members that attend meetings often.

Fortunately the survey includes a question which determines the frequency
with which the member attends meetings.   The prevalence of such a member
in the population sampled is inversely proportional to this frequency.

When this correction is applied to determine the average frequency of
meeting attendance in the population sampled, a significantly different
result from the sample average is obtained.   For example, the sample
average frequency was more than 4 meetings per week in 1989, but the
population average turns out to be 2.9 meetings per week.

This population average is obtained by correcting, for frequency, meeting
frequency data for each of 22 age groupings and 7 ranges of frequency.

In principle the same correction should be applied to all attributes
determined by the survey.   However, the detail to which the data is
reported makes this easy only for the 22 age groups, not for the seven
levels of frequency.  As an example, data on the sex of 19- and 20-year-
olds can be corrected for the average frequency of attendance at meetings,
but if there were a cross-correlation of meeting frequency with sex of the
respondent for that age group, the correction would not take it into
account.

Fortunately there appears to be little variation of average frequency with
age group except for very young members, with the result that the sample
results for almost all sttributes other than frequency itself and drug
addiction appear to be representative of the sample population (see
Appendix E).  This has been verified for all the attributes cited in this
report for the 1989 and 1986 surveys, and has been verified as well for
many of the attributes of earlier surveys.

Therefore, comparison from one survey to another in Appendix B, for the
purpose of identifying trends, is made using sample averages, which are
easily accessible in earlier surveys.

-----BEGIN PGP SIGNATURE-----
Version: PGP Personal Privacy 6.5.8

iQA/AwUBQk84Ebw9MOKEeRC8EQKQQACgqSFiUVunAUBdqRfqDicOQwNbhtUAnRAb
WsYifSquEMX9Pm9btISkfwrG
=xgXl
-----END PGP SIGNATURE-----