MRI To Increase Sample To 30,000
MRI's national study sample, 20,000 adult respondents annually, is
one of the largest in media and advertising research. Nonetheless,
over the years and especially recently, many MRI clients have
expressed interest in our increasing the size of the sample to enable
still more reliable reporting and analysis of MRI data. The
"fragmentation" of media audiences and the proliferation of new media
and brands have fueled this demand.
MRI, recognizing a larger sample would indeed yield greater data
reliability and enhance the confidence with which the data are used,
is as eager as our clients to increase the size of the sample.
MRI's annual sample size will be increased to 30,000 respondents.
This increase will yield enhanced reliability of overall media
audience estimates, and also of estimates based on target audiences
and other subgroups. On average, the sample size increase will lower
sampling error by 18 percent. This improved reliability will be
achieved in each spring and fall release of a year's data and in each
release of DoubleBase (two years' data).
MRI has considered modifying the allocation of elements of its
sample as the total sample is increased; we have subjected the
present sampling plan to careful scrutiny and evaluation of its
composition by income strata, by geography, and by other variables.
We have been gratified to conclude that no change in the sampling
plan would yield an increase in either effectiveness or efficiency,
and accordingly plan simply to add sample clusters proportionate to
the composition of the present sample. This will ensure consistency
with previous data collection.
It is estimated that full implementation of this increase will
require four years of incremental additions of replicates.
Recruiting, training, field-testing, evaluating, deploying, and
controlling additional interviewers must be accomplished
deliberately. To compromise data collection quality to attain an
earlier increase in total sample size would be unacceptable.
Sample will be added in each of eight consecutive data collection
periods (waves) beginning with Wave 37, which goes into the field in
early March, 1997. Alternate waves will add sample from major markets
and smaller markets; this enables close oversight of the increases
while enhancing reliability of more- and less-urban areas in tandem
(but with no expected effects on estimates).
MRI's goal in this matter is to be responsive to the needs of its
clients. To be able to do so and at the same time improve the
reliability and value of our research is, we believe, a great
opportunity for all concerned.
Back to Top
ENHANCED ASCRIPTION
MRI is researching and developing enhancements to its procedures
for ascribing missing data. The enhanced procedures offer the promise
of improving the matching of respondents to non-respondents.
Basic Ascription:
Ascription is a process by which survey respondents who have not
answered all survey questions are assigned responses for these
"missing items." The MRI study is conducted in two phases. The first
part, a personal interview, measures magazine audiences, other media
usage, and demographic characteristics. The second part is a
self-administered product information booklet (PIB) which is
completed by more than 60 percent of the personal interview
respondents. To complete the data records of non-PIB-returning
respondents, MRI's computerized ascription program pinpoints the best
match between a product booklet non-respondent and a product booklet
respondent by comparing multiple demographic and behavioral
variables. The missing data in the non-responder's booklet is then
filled in with the full PIB data record of his/her matched responder.
Although ascription is not perfect, its benefits are substantial:
- By ascribing product booklet responses, all respondents' media
data are retained, thus preserving the high response rates (70
percent +) from the personal interview.
- Ascription preserves the product consumption levels among
responders and within the referenced (matched) demographic
subgroups.
- Since the records of personal product usage and household
product usage are ascribed in their entirety, between-product
correlations are preserved.
- An analysis of two years of our data shows ascription retains,
in an overwhelming majority of cases, actual media consumption and
product usage levels.
The results produced by ascription also adequately preserve
complex relationships among magazines and other media collected in
the personal interview and product usage and other behaviors
collected in the PIB. Even in the relatively rare situations in which
relationships are affected by the current ascription procedure, they
are at worst attenuated or diminished, not obscured.
Enhanced Ascription:
MRI is developing a computer-intensive approach to address
situations in which basic ascription does not fully preserve observed
relationships between personal interview data and information
collected in the PIB. Enhanced ascription entails the computer
searching for situations in which the basic ascription process does
not fully preserve a relationship between two variables: the software
identifies instances of basic ascription resulting in attenuation of
relationships.
- First, enhanced ascription weights the respondents with
completed PIBs to produce a "weighted-recovered-only" estimate.
- Second, basic ascription is conducted for respondents who did
not complete a product booklet. The total sample -- completes and
ascribed -- is tabulated, using full sample weights, to produce
"basic ascribed" estimates.
- Third, the enhanced ascription software searches for
situations in which the "weighted-recovered-only" estimate differs
from the "basic ascribed" estimate. The software examines all
magazines and other personal-interview-measured media against all
products and behavior data collected in the PIB. The results of
these searches are sets of media-product pairs.
- Finally, once these pairs have been identified, the computer
algorithm applies enhanced ascription optimized for each divergent
pair. Going beyond basic ascription, enhanced ascription is
directed specifically at dealing optimally with the residual
attenuation associated with specific magazine-product pairs.
It is important to note that enhanced ascription does not affect
"random" variation; rather it is used only to improve basic
ascription when significant attenuation occurs.
Enhanced ascription has been under development since August, 1996.
Research and development is scheduled to be completed by the end of
1997. We will report our progress to the industry at the Worldwide
Readership Research Symposium in Vancouver in October, 1997, and to
our clients as developments occur.
Back to Top
THE TRUTH ABOUT CATS' AND DOGS' (OWNERS)
Owners of dogs and cats are different animals, according to
Mediamark data. Nearly three in ten (29 percent) American adults own
at least one dog, 22 percent own a cat, and one in ten own both a cat
and a dog.
"We'll get a pet when you're old
enough to take care of it."
The presence of children aged 12-17 is a distinguishing
characteristic of dog and cat owners. Families with children this age
owning both a dog and a cat index 170 against the total population,
families owning a dog with the presence of children 12-17 have an
index of 139, and cat owners with teens have an index of 130.
Although they are more likely than the general population to own a
dog or a cat, families with (younger) children are less likely than
those with older kids to have these pets.
Dog owners (118 index) are more likely than cat owners (110 index)
to be married.
Southern Country Dogs:
Cat owners are more likely (138 index) to live in New
England, while dog owners (122 index) are more likely to live in the
South. While dog owners
(121) are slightly more likely than cat owners (118) to live in rural
areas (D counties), owners of both a cat and a dog index at 163 for
living in the country.
Dog owners (118 index) are slightly more likely than cat owners
(111) to be homeowners rather than renters.
Media Animals:
Dog owners are likely to be heavy radio listeners (115
index), but are not as likely to be heavy readers of magazines or
newspapers or heavy viewers of primetime TV. (With lower newspaper
readership, how do they train their dogs?)
On the other paw, cat owners are more likely to be heavy magazine
readers (116 index) and heavy radio listeners (120 index).
AV-CASI Software In Field Testing
Mediamark's Audio/Visual Computer- Assisted Self-Interviewing technology,
AV-CASI, is now in field test. The February/March test is being conducted among
respondents in clusters matching sites in which paper-and-pencil interviews are also
being conducted. The AV-CASI technology uses a portable computer with a
touch-sensitive screen to collect the MRI personal interview data. A vocal prompt from
the computer guides the respondent and reinforces the stimuli as they appear on the
screen.
Testing and refinement of the software have been ongoing since 1995 with results to date
demon-strating AV-CASI can ensure superior quality control of fieldwork and more
efficient data collection. 'We are optimistic AV-CASI will improve and enhance the
collection of respondent data, and we look forward to the results of our field test,"
comments Mediamark's Marty Frankel.
Back to Top
Mediamark Reporter is a tool to analyze and report all data
formerly distributed through printed volumes. It is available to all
MRI clients at no charge.
