Next: » URBANISM NOTE
Up: anes_cdf_app
Previous: » 1992 ERRORS IN
NOTE ON USE OF WEIGHTS:
~~~~~~~~~~~~~~~~~~~~~~
THERE ARE 6 WEIGHT VARIABLES IN THE CUMULATIVE DATA FILE:
VCF0009 VCF0010 VCF0011
VCF0009a VCF0010a VCF0011a
1. IMPORTANT: CASES FROM YEARS OTHER THAN 1970 AND 1994-LATER
HAVE IDENTICAL VALUES IN ALL (6) WEIGHT VARIABLES. WHEN DATA
FOR YEARS EXCLUDING 1970 AND 1994-LATER ARE USED FROM THE
CUMULATIVE DATA FILE, ANY OF THESE WEIGHT VARIABLES MAY BE
UTILIZED TO OBTAIN THE SAME RESULTS.
2. WHEN CUMULATIVE DATA FILE VARIABLES INCLUDE 1970 DATA AND/OR
DATA FROM 1994 AND LATER, WEIGHT SELECTION SHOULD BE BASED
UPON 2 CONDITIONS:
A) THE 'TYPE' OF 1970 VARIABLE
B) WHETHER POST-STRATIFIED OR NON-POST-STRATIFIED WEIGHTING
IS USED 1994 AND LATER
2a.
NOTES CONCERNING 1970
---------------------
The original study data for 1970 were weighted by variable
type; there were 3 types of 1970 variables: Type 0, Type 1,
Type 2.
Most 1970 variables were Type 0. 1970 Type 1 or 1970 Type 2
data are identified in documentation; all other Cumulative Data
File variables that include 1970 data and not specifying otherwise
are assumed to be Type 0 data for 1970.
Weighting Cumulative Data File variables that include 1970
data means selecting the appropriate weight for the type of
1970 variable represented in the variable.
For each of the 3 types of 1970 variables, there are 2
Cumulative Data File weight variables which have identical
values for 1970 cases:
CDF VAR CDF VAR CDF VAR
INCLUDES 1970 INCLUDES 1970 INCLUDES 1970
TYPE 0 DATA - TYPE 1 DATA - TYPE 2 DATA -
USE ONE OF: USE ONE OF: USE ONE OF:
~~~~~~~~~~~ ~~~~~~~~~~~ ~~~~~~~~~~~
VCF0009 VCF0010 VCF0011
VCF0009a VCF0010a VCF0011a
If no 1994 or later cases are selected in combination with 1970
cases, then it does not matter which of the 2 weight variables
for that 'type' is used. For example: if 1970 data are Type 0
and no 1994 or later data are included, then either VCF0009 or
VCF0009a can be used to produce the same results. Choice among
the 2 variables for each 1970 'type' is determined by the kind
of weighting desired for years 1994 and later (2b).
2b.
NOTES FOR 1994 AND LATER (Post-stratified weights available)
------------------------------------------------------------
1994 and later data have more than one type of weighting
available:
1994-LATER
WEIGHTING TYPE WEIGHT VALUES SAME IN 3 VARS:
~~~~~~~~~~~~~~ ~~~~~~~~~~~~~~~~~~~~~~~~~~~~
1. NOT POST-STRATIFIED VCF0009, VCF0010, VCF0011
2. POST-STRATIFIED VCF0009a, VCF0010a, VCF0011a
In Pre-Post studies after 1994, there were 2 post-stratified
weights, one for pre-election variables from Pre-Post studies
and one for post-election variables from Pre-Post studies.
Cumulative Data File "a" weight variables (VCF0009a,VCF0010a,
VCF0011a -- identical to one another for these years) correspond
to the 'pre' post-stratified weights from these years.
Note: although the post-election post-stratified weights provide
the most complete adjustment for post-election variables from
pre-post studies after 1994, the difference in estimates produced
using pre-election post-stratified weighting versus post-election
post-stratified weighting for post-election variables is small
enough to justify using pre-election post-stratified weights
(VCF0009a,VCF0010a,VCF0011a) for all data when post-stratification
adjustments are desired.
ADDITIONAL NOTES
~~~~~~~~~~~~~~~~
1970 notes:
Black respondents were asked all questions. White respondents
were assigned to one of two formats of the questionnaire. Each
format of questionnaire excluded a selection of questions if the
respondent was white; the questions excluded differed in each
format. The two formats of the questionnaire were designated
Form 1 and Form 2.
'Type 0' describes variables for questions which were not excluded
in either questionnaire format. 'Type 1' describes variables for
questions excluded from Form 2, and 'Type 2' describes variables
for questions excluded from Form 1.
All 1970 vars not listed in VCF0010 as type 1 or in VCF0011 as type
2 are type 0; documentation for each Cumulative Data File variable
in this codebook provides the type of 1970 data utilized (if any).
1994 and later notes:
The first time-series weight variables produced for the 1994 and
1996 studies did not correct for an under-representation in Study
data of younger and less-educated respondents (primarily an effect
of the empaneling design used 1992 to 1996; see the 1994 and 1996
codebook special weight documentation). New weight variables were
therefore produced, constructed to post-stratify the NES data to
match the Current Population Study (CPS) estimate of the
distribution of age group by education level. After 1996, the
same type of post-stratified weights continued to be constructed.
Walter Mebane
2005-11-03