Note to Macintosh Users: To see this page correctly, you need
to change your Font Encodings in Netscape/Explorer to MacRoman
Statistical Consulting Center Annual Report
Glen H. Laird
May 28 1998
1 INTRODUCTION
The Statistical Consulting Center (SCC) at Florida State University is
a research assistance facility for the students, faculty, and staff at
FSU. While clients outside of FSU are sometimes charged a fee, the consulting
center is completely free of charge for FSU students, faculty, and staff.
We are staffed by one or more experienced graduate students with faculty
oversight. Services include, but are not limited to:
Translating hypotheses into statistical terms
Designing sampling procedures
Choosing appropriate statistical methods
Interpreting computer output
Phrasing statistical results
Referrals to other statistical help
The Consulting Center generally does not perform actual analyses. However,
clients are free to reschedule further consultation appointments if an
initial visit is insufficient.
My appointment hours for summer of 1998 are Tuesday from 2 P.M. to 4
P.M., Wednesday from 10:30 A.M. to 12:30 P.M., and Thursday from 9:30 A.M.
to 11:30 A.M. However, if those hours are inconveninent, arrangements can
sometimes be made for different hours or another consultant. Appointments
are generally made for two hour blocks, but the entire time does not have
to be used.
If you are a potential client and believe you may need statistical assistance,
we recommend getting assistance at the earliest possible stage of your
research. For more on how to make an appointment, Consulting Center policy,
and FAQ, call the main office of the FSU Statistics Department at 644-3218
or visit our website at http://stat.fsu.edu/consult/index.php.
I can be reached at 644-5755 or laird@stat.fsu.edu
.
2 SUMMARY OF BUSINESS ACTIVITIES
In the past year, I have seen approximately 65 separate clients in person.
Approximately 10 additional clients were handled by phone, email or FAX.
Most of our clients, perhaps 85%, were graduate students at FSU working
on their thesis or dissertation. About 10% were faculty at FSU. Most of
the rest were from outside of FSU, although at least one undergraduate
was also assisted. There was also overlap in the above categories, such
as FSU faculty members working on a degree from another university. Clients
from FSU came from a wide variety of schools/departments including:
Anthropology
Biology
Business
Education
Meteorology
Music
Nursing
Oceanography
Physical Education
Psychology
Social Work
Textiles and Consumer Sciences
FSU-FAMU College of Engineering
In addition, clients from other FSU organizations, or from outside FSU
were assisted, such as:
Florida Deparment of Labor
Florida Department of Environmental Protection
FSU Varsity Volleyball coach
FSU library system
FSU Police Department
Seminole Boosters
Typically, clients were seen about twice, although this varied by the client's
need. Some clients were seen four or five times. One visit clients were
not uncommon either.
A number of different services were requested by clients. Many needed
help designing a survey or sampling scheme, including issues of reliability,
validity, sample size, and power. About half had already collected their
data and were more concerned with choice of statistical procedure or interpreting
and phrasing statistical results. For some clients, I read their results
to see if they were accurate and clear. One client had no data at all,
but instead had questions about distribution theory.
Clients also had a wide variety of statistical backgrounds. Some clients
were unfamiliar with even basic terminology used in statistics. Others
had knowledge of specific statistical models that required me to learn
additional background information to fully understand the client's research.
Most clients had a statistical background somewhere in between those two
extremes.
3 TWO TYPICAL CASES
The following cases may be illustrative of the kind of work done in the
SCC. They can be considered somewhat typical, though still noteworthy enough
to be mentioned separately.
3.1 CASE ONE
I received an interesting case from a graduate student in the Physical
Education department. He was studying the effect a sports drink (Powerade)
on the length of time one could run on a treadmill. He arranged for 8 men
and 8 women to participate in his study. Sample sizes as small as these
are common in drink comparisons.
He measured a number of variables, but was mainly interested in whether
the subjects could run longer when drinking Powerade than when drinking
a placebo. He had to stop measurement, for ethical reasons, upon anyone
running for more than 100 minutes. He therefore had a survival analysis
situation with censored data. The unusual aspect of his case was that he
had the same subject take both treatments on two separate weekends. His
experiment was therefore paired. Survival analysis experiments are usually
not paired because when the subject reaches his/her failure time for one
treatment, he/she is incapable of participating in any further study. For
example, in cancer research, when a patient in one treatment group dies,
the patient is no longer alive to participate in the study as a member
of any other treatment group.
This case required me to go to the library and look up some references
on the analysis of paired survival data, which was an unfamiliar topic
to me. I ended up finding a journal article recommending the paired Prentice-Wilcoxon
statistic in this situation. Unfortunately, I have not yet heard back the
results of this test on my client's data, but I feel I have already benefited
from the additional statistical knowledge gained.
3.2 CASE TWO
I received an equally satisfying case from a student in the Education department.
He was studying the effect of various activities in which college students
participate on ``wellness.'' Briefly put, wellness is an abstract concept
measuring a person's overall health; physically, mentally, and spiritually.
He administered his survey to about 300 students in different universities
in the southeastern U.S. He asked students about the number of hours they
spent doing various activities such as homework, partying, or sleeping.
He also recorded some demographic variables.
The methods used to analyze his data were fairly standard. We used a
stepwise regression model. However, the client's knowledge of statistics
was rather limited. I explained a number of statistical concepts to him,
such as variable interaction, strength of association, and p-values. Over
several sessions, he gradually came to understand the concepts quite well.
It was this interpersonal aspect of the case that was most rewarding to
me. After the regression had been run, we found that people who were the
most ``well'' tended to be people that spent more time in fitness and religious
activities, and spent less time alone. These results may seem ``obvious'',
but it is easy to think of several other ``obvious'' results (such as differences
by race) that did not occur.
This case also involved a data quality issue since a few of the students
put down more than 168 hours of activities in one week, which is impossible.
I believe the client learned some practical lessons about how subjects
fill out questionnaires.
4 REFLECTIONS
I would like to say that I have enjoyed working in the SCC very much this
year, and I look forward very much to working with future clients. I need
the statistical learning experience of working with ``real'' data not from
a textbook, and I enjoy the interaction with clients on both a professional
and personal level. I think I've learned a little about everything from
ants to education to hurricane damage.
I would also like to thank a number of professors in the FSU statistics
department, especially Dr. Duane Meeter, Dr. Pi-erh Lin, Dr. Ian McKeague,
and Dr. Xufeng Niu, for assistance with some unusual or difficult cases.
Furthermore, I appreciate Dr. Zahn's generosity in loaning me a copy of
the Sage Monograph series.
|
