Activity Number:
|
428
|
Type:
|
Contributed
|
Date/Time:
|
Wednesday, August 9, 2006 : 10:30 AM to 12:20 PM
|
Sponsor:
|
Section on Statistics in Epidemiology
|
Abstract - #306855 |
Title:
|
Comparing Methods of Examining Trend Data for a Large Population
|
Author(s):
|
Kristen Eberly*+ and Barbara Neas and David M. Thompson
|
Companies:
|
The University of Oklahoma and The University of Oklahoma Health Sciences Center and The University of Oklahoma
|
Address:
|
P.O. Box 26901, CHB309, Oklahoma City, OK, 73190,
|
Keywords:
|
trend ; graphical comparison ; large sample size ; confidence interval ; moving average ; test of trend
|
Abstract:
|
The objective of this study is to compare methods of evaluating time trend data to determine a meaningful trend. Deaths from lung cancer and related diseases in OK from 1980 to 2004 are shown for the entire population as well as stratified by sex and age. Evaluation of trends in large datasets can often produce statistically significant results due to the sample size. With the use of traditional alpha, non-meaningful trends are almost always statistically significant. Common methods such as confidence intervals, moving averages, and trend tests have been shown to be statistically significant. When looking at short time intervals compared to long intervals, statistical significance is still achieved when the actual trend may not be considered practically significant. In this study, graphical comparisons demonstrate the most meaningful interpretations.
|