This page contains a Flash digital edition of a book.
research career. (After coming to NCHS, race—estimates of the counts that would types (continuous, categorical, count) were
I also became an adjunct professor in the have been obtained had the prior standards used as predictors, but they sometimes had
Joint Program in Survey Methodology at been in effect (www.cdc.gov/nchs/about/ small amounts of missingness as well.
the University of Maryland, so I didn’t have major/dvs/popbridge/popbridge.htm). The I was involved in many aspects of this
to abandon academia completely.) Finally, bridging process used models predicting
project (as with the bridging project), but
the Washington area has an exceptionally responses under the prior standards from
especially in developing and evaluating the
active statistical community. For example, responses under the current standards and
methods for imputation, writing technical
it is home to the largest chapter of the ASA: covariates. These bridging models were
documentation, and consulting with ana-
the Washington Statistical Society. developed using data from NCHS’ National
lysts using the multiply imputed data.
Do I miss being a full-time professor at a Health Interview Survey (NHIS), which has
fine university? Yes. Am I glad that I changed allowed multiple-race responses for many
An Exciting, yet Sometimes
jobs? Absolutely. Besides the fact that years, but which also asks multiple-race
‘Scary,’ Job
changing jobs every once in a while helps to reporters for a single race group that best
The two projects just described have sever-
'keep one’s blood flowing,' I’ve enjoyed the describes the person in question.
al general characteristics in common. First,
atmosphere and work at NCHS. I’m less of I worked on many aspects of this project,
they required teamwork, both among staff
a free agent in some ways now than I was but perhaps my biggest contributions were
at NCHS and between NCHS and other
in academia (although I also don’t have to to the formulation of the overall approach
organizations. Second, they required sta-
hustle for grant support the way I did), and used to solve the problem, the development
tistical research. In fact, both projects led
there are some bureaucratic hassles involved of the bridging models, and the derivation
to publications in JASA [98(464) and
in working for the government; but every of methods for assessing uncertainty in anal-
101(475)]. Third, the problems were
job has its own administrative headaches, yses that use the bridged data.
complicated, and they required making
and any annoyances due to governmental A second project involved multiple
assumptions and implementing approxi-
red tape have been far outweighed by the imputation of missing income data. The
mations for their solution. Finally, there
opportunities I’ve had for interesting work. NHIS provides a rich source of data for
was great interest in the data, and people
studying relationships between income
both inside and outside the agency were
A Small Sample of My Projects
and health and for monitoring health
clamoring for release of the data while the
To illustrate some of the statistical work and health care for persons at different
work was going on.
that goes on at NCHS, I’ll briefly outline income levels. However, the nonresponse
Projects with the second through fourth
a couple of major projects in which I’ve
characteristics create the need to “push the
rates are high (roughly 30%) for two key
been involved.
envelope” with regard to methodology
items: total family income in the previous
In 1997, the Office of Management
under time constraints while maintaining
calendar year and personal earnings from
and Budget revised the standards for clas- employment in the previous calendar year.
the integrity of the data. This results in a
sifying federal data on race and ethnicity. To handle this missing-data problem and
job that is exciting, yet sometimes ‘scary.’
A key provision was that respondents to allow analysts of the data to assess the
Challenges and Opportunities
federal data collections be given the option uncertainty due to missing data, mul-
Statistical work at NCHS, and in the
of choosing more than one race group to tiple imputation of these items has been
government in general, is challenging
describe the person in question. Because the performed for several years (www.cdc.gov/
for administrative and technical reasons.
previous (1977) standards called for only a nchs/about/major/nhis/2006imputedincome.
single race group to be reported, data col-
For example, budgets are tight; costs are
htm) in collaboration with researchers at
lected under the 1977 standards are not
increasing for data collection, processing,
the University of Michigan.
comparable with data collected under the
and dissemination; there is a desire on the
Several features of the data made the
revised standards. This can cause problems project particularly challenging and inter-
part of policymakers, researchers, and oth-
in trend analysis, the calculation of a vital
ers for more information; and the protec-
esting. First, the data are hierarchical
rate—for which the number of events in
tion of confidentiality is becoming more
in nature, with one of the key variables
difficult. Oh, yes, and let’s not forget the
the numerator and the population size in reported at the family level and the other
famous “graying of the federal work force,”
the denominator can come from different reported at the person level. Second, there
with many employees nearing retirement
sources using different standards—and in were cases in which the value of one vari-
age.
other types of studies that combine data able (e.g., personal earnings) could be
Of course, with challenges come
classified under the two standards. restricted by the value of another variable
opportunities. Difficulties due to sparse
The decennial census, a widely used (e.g., whether the person was employed),
resources and conflicting desires and con-
source of denominators for vital rates, began but the values of both variables were miss-
straints with regard to data will need to be
allowing multiple-race reporting in 2000. To ing simultaneously. Third, in some cases,
addressed through the creative develop-
make the 2000 census data (and intercensal family income and personal earnings need-
ment of efficient methods for collecting
and postcensal estimates) comparable ed to be imputed within bounds because
new data and methods for getting the most
with data classified according to the 1977 partial information was available about
out of analyses of existing data. And the
standards, NCHS, with assistance from the them (e.g., when a range was provided for
‘graying’ work force will result in openings
U.S. Census Bureau, produced “bridged” family income, rather than an exact dollar
for new generations of statisticians to carry
census counts by county, age, sex, and
value). Finally, several variables of various out such work. ■
SEPTEMBER 2008 AMSTAT NEWS 3
SEPTEMBER AMSTAT FINAL.indd 3 8/20/08 2:26:54 PM
Page 1  |  Page 2  |  Page 3  |  Page 4  |  Page 5  |  Page 6  |  Page 7  |  Page 8  |  Page 9  |  Page 10  |  Page 11  |  Page 12  |  Page 13  |  Page 14  |  Page 15  |  Page 16  |  Page 17  |  Page 18  |  Page 19  |  Page 20  |  Page 21  |  Page 22  |  Page 23  |  Page 24  |  Page 25  |  Page 26  |  Page 27  |  Page 28  |  Page 29  |  Page 30  |  Page 31  |  Page 32  |  Page 33  |  Page 34  |  Page 35  |  Page 36  |  Page 37  |  Page 38  |  Page 39  |  Page 40  |  Page 41  |  Page 42  |  Page 43  |  Page 44  |  Page 45  |  Page 46  |  Page 47  |  Page 48  |  Page 49  |  Page 50  |  Page 51  |  Page 52  |  Page 53  |  Page 54  |  Page 55  |  Page 56  |  Page 57  |  Page 58  |  Page 59  |  Page 60  |  Page 61  |  Page 62  |  Page 63  |  Page 64  |  Page 65  |  Page 66  |  Page 67  |  Page 68  |  Page 69  |  Page 70  |  Page 71  |  Page 72  |  Page 73  |  Page 74  |  Page 75  |  Page 76  |  Page 77  |  Page 78  |  Page 79  |  Page 80  |  Page 81  |  Page 82  |  Page 83  |  Page 84
Produced with Yudu - www.yudu.com