Transcript Statistical Consulting Lab

Statistical Consulting Laboratory
(SCL)
Website: http://academics.utep.edu/statlab
Border Biomedical Research Center
(BBRC)
SCL Personnel
 Dr.
Peter Moschopoulos (Core Director)
 Dr. Joan Staniswalis
 Dr. Ming-Ying Leung (Bioinformatics)
 Dr. Ori Rosen
 Dr. Naijun Sha
 Dr. Julia Bader: (Full-Time Consultant)
 Mr. Yash Dayal, M.S. (Network Manager)
 Ms. Bonnie Smith (Administrative Secretary)
Statistical Consulting Laboratory
http://academics.utep.edu/statlab
Mission
The mission of SCL is to provide statistical support to researchers in
the BBRC and become a regional resource center for the statistical
support of research in basic sciences, health sciences, education,
medical sciences and other fields.
Statistical Consulting Laboratory
http://academics.utep.edu/statlab
BBRC Present Funding

Support for Personnel:






Equipment Upgrades



Statistician (Dr. Julia Bader) 12 months at 50%
SCL Director 1 summer month at 100%
Secretary: 12 months at 100%
Network Manager: 12 months at 30%
Resource Statistician 1 summer month at 100%
Upgrades of Sun workstations and PC Network
Travel, Materials, Supplies and Statistical Software Manuals
Other Direct Costs

Software licenses such as: SAS, S-plus, Current Index to Statistics, etc
Statistical Consulting Laboratory
http://academics.utep.edu/statlab
SCL Activities

The SCL provides statistical data analysis support to the
BBRC participants. This includes help in design of
experiments, power calculations and assistance in grant
proposal preparation.

The SCL fosters collaborations between SCL statisticians
and Researchers at UTEP who need statistics in their
research.

The SCL offers fee-based Statistical Consulting.
Statistical Consulting Laboratory
http://academics.utep.edu/statlab
BBRC Consultations (2005-2006)

Dr. William Baldwin:
Statistical support, including power calculations in preparation of a
research proposal to NIH, and statistical analyses for two manuscripts.

Dr. Elizabeth Walsh:
Statistical analyses for studies of rotifer palatability to invertebrates.

Dr. Renato Aguilera and Dr. Armando Valera:
Collaboration on a study of HIV infection and risk behavior of Hispanic
farm workers.

Dr. Sha and Dr. Zhang:
Developed and taught (twice) a class of post – genomics analysis for
the Bioinformatics Program.
Statistical Consulting Laboratory
http://academics.utep.edu/statlab
BBRC Consultations(2005-2006)

Dr. Joanne Ellzey:
Statistical analyses for two related arsenic studies on mice.

Drs. Das, Sha and Bader:
Experimental design, sample sizes and statistical analysis on a
microarray study on colon cancer.

Dr. Zhang, Dr. Sha, Dr. Bader:
Study of auto antibodies in liver cancer patients. Funded by the Center
for Border Health Research; also an NIH proposal on identifying novel
cancer-related antibody-antigens in lung cancer study.
Statistical Consulting Laboratory
http://academics.utep.edu/statlab
About Power of Tests

Very General Setting:
The researcher tests a claim that in fact is true or false, e.g. that cancer is
present. This claim is to be rejected or retained on the evidence of data.
There are two types of errors made, Type-I and Type-II .
Type-I - Probability of rejecting the claim/claim is true
 Type-II - Probability of accepting claim/claim is false

Usually we control Type-I at .01 or .05.
Type-II depends on the sample sizes. We choose the sample sizes so
that Type-II is minimized.
Power = 1 – Type-II = Probability of rejecting the claim when it is in fact
false. Usually we require power of 80% or higher and calculate the sample
sizes to achieve it according to the wishes of the investigator.
Statistical Consulting Laboratory
http://academics.utep.edu/statlab
Example:
Sample size for each group for a t-test for the difference in means (Treated
versus Control), testing mean difference = 0 (versus ≠ 0 (two-sided) or versus
> 0 (one-sided)), alternative mean difference = minimum difference, α = 0.05
and 80% power with varying effect size (minimum difference)/(standard
deviation).
Effect Size
Two-Sided
One-Sided
0.2 (small)
394
310
0.3
176
139
0.4
100
78
0.5 (medium)
63
51
0.6
45
36
0.7
34
26
0.8 (large)
26
21
Statistical Consulting Laboratory
http://academics.utep.edu/statlab
Dr. William Baldwin (Power Calculations)


Sample size justification for proposal with mice and human cell
lines, response variable is the induction rate
Sample power table for repeated measures design for comparing
3 groups (1 (Control), 2, 3) for human hepatocyte cell lines
Means
1vs2–1vs3
Std.
Dev.
Overall
n=8
1 vs 2
n=8
1 vs 3
n=8
Overall
n=10
1 vs 2
n=10
1 vs 3
n=10
2-4 fold
2 fold
.39
.15
.79
.49
.18
.89
2-4 fold
3 fold
.19
.10
.46
.24
.11
.56
4-8 fold
2 fold
.98
.79
1.00
.99
.89
1.00
4-8 fold
3 fold
.75
.46
.99
.87
.56
1.00
Statistical Consulting Laboratory
http://academics.utep.edu/statlab
Dr. Liz Walsh
Study of Rotifer Palatability to Invertebrates
3 factor feeding study

Predator








0 hours
12 hours
18 hours
24 hours
Prey




Control (None)
Damselfly Nymph
Dragonfly Nymph
Hydra
Time


Figure. Percent Survival for Colonies
and E.senta
100
Percent survival

80
Cont rol E.sent a
Damself ly E.sent a
60
Dragonf ly E.sent a
Hydra E.sent a
Cont rol Colonies
40
Damself ly Colonies
Dragonf ly Colonies
Hydra Colonies
20
0
Colonies
E. senta
0
6
12
18
24
Time (hour)
Response variable as the percent surviving out of 5
Repeated measures over time, using the General Linear Mixed Model analysis
showed predator*time*prey interaction significant (p = 0.0029), with Tukey post hoc
procedure for comparison of means
Statistical Consulting Laboratory
http://academics.utep.edu/statlab
Present Clients (2005)






WBAMC continuing since Oct. 1, 2003
Dr. Arthur Blume, UTEP Promoting Health Parity among
Mexican-American Women funded by NIH.
Center for Institutional Evaluation, Research and Planning
(CIERP).
College of Health Sciences (all grants)
Center for Border Health Research (reviewing grant
proposals)
Various UTEP and Off-Campus Researchers
Statistical Consulting Laboratory
http://academics.utep.edu/statlab
Goals




Continue support for BBRC, increase participation in grant
proposals.
Participate in design of experiments with animals in the new
research facility.
Increase visibility of SCL across UTEP and the Border Region.
(brochure preparation underway, website ready, marketing
strategies considered, NIH competitive grant application, COPI with Dr. Ibarra, College of Health Sciences).
Continue with current projects, WBAMC, CIERP, College of
Health Sciences and CBHR support of grant proposals.
Statistical Consulting Laboratory
http://academics.utep.edu/statlab
Questions?
Statistical Consulting Laboratory
http://academics.utep.edu/statlab