The 4th Annual

NC State University

Undergraduate Summer Research Symposium

 

Statistics VIGRE Traineeship

 


Abstracts are listed in alphabetical order by the last name of the corresponding author.

 

 

 


 

 
Student Author(s): 

Chaudoir, Kristin M.

Misenheimer, Don M.

Home Institution:

North Carolina State University

Program:

Statistics VIGRE Traineeship

Department(s):

Statistics

Research Mentor(s)

Marcia L. Gumpertz/Statistics

David T. Butry/ Statistics

Title of Presentation:

Exploring Lightning-Caused Wildfires

 

 

Each year, wildfires destroy millions of acres of land in the United States. Over 30% of ignitions in our research area, the St. Johns River Water Management District of Northeastern Florida, were caused by cloud-to-ground lightning strikes.  The purpose of our exploratory analysis of the lightning and wildfire data is to further understand the relationship between lightning characteristics and wildfires. 

            Using a regression model with lightning and fire data from June and July 1998, it can be shown that the total number of fires in an area is related to several of the lightning characteristics. 

            The most prominent finding, so far, has been the possible discrepancies we have found among the data.  Fire cause, location, and ignition time are extremely hard to pinpoint. Fires can smolder for weeks or months before spreading into a noticeable fire.  Lightning detection also presents a location accuracy problem.  The strokes can touch ground several kilometers away from one another.  This makes it difficult to identify the precise location where a lightning strike caused a fire.

 

 


 

 
Student Author(s): 
Gracien, Katina F.
Hare,  Brian 
dsHome Institution:
North Carolina State University
Program:

Statistics VIGRE Traineeship

Department(s):
Statistics
Research Mentor(s)
William F. Hunt/Statistics
Title of Presentation:

What Can Be Done to Improve the Accuracy of Fine Particulate Matter Reported to the Public Using the Air Quality Index (AQI)?

 

 

PM 2.5 is particulate matter that is 2.5 micrometers or less in aerodynamic diameter and is caused by various processes that emit particles into the air.  These particles, over time, have been linked to cardiovascular and respiratory diseases especially in younger and older individuals.  Thus there is a large concern of how well the people that are being exposed to various amounts of this pollutant are being informed about it so they can take the necessary steps to protect themselves.  The EPA’s AQI (Air Quality Index) entails 6 levels of health concern that are provided for citizens via television and newspaper.  Two databases were examined in this study, the Air Quality System (AQS) database consisting of quality assured Federal Reference Method fine PM data, and the AirNow database which consists of continuous fine PM measurements provided to EPA on a real time basis used to predict the AQI. Comprehensive exploratory statistical analyses and visualizations were conducted to examine the causes of the variation between the AQS and the AirNow databases.  Discrepancies were found in the concentrations reported to AirNow and AQS. On some occasions the AirNow database reported 2-3 AQI levels above or below what was reported to the AQS database.  The various locations and seasons of where and when these discrepancies in the AirNow occurred were found.  These findings suggest a difference in the way the AirNow values and AQS values in the various states are being reported to the EPA.  Developing a quality control procedure to pair the “problem sites” with sites that consistently report data accurately may lead to an early warning system to detect data that has not been properly adjusted at a “problem” site.  It is critically important that the general public be accurately informed regarding air quality so they can properly plan their daily activities.

 

 
 
 
 
 
 
 



 
 
 
 
 
 
 
 
Student Author(s): 

Hare, Brian C.

Gracien, Katina F.

Home Institution:

North Carolina State University

Program:

Statistics VIGRE Traineeship

Department(s):

Statistics

Research Mentor(s)

William F. Hunt/Statistics

Title of Presentation:

Creating an Emission Standard Using a

Statistical Approach

 

 

Why do we need emission standards?  We need them to help keep industries from producing excessive amounts of harmful chemicals.  These standards are enforced by a regulating company that monitors the chemicals emitted from pollution sources.  To help monitor these pollution levels, there is usually a limit or emission standard that cannot be exceeded.  These can be in the form of a rate or a maximum concentration amount.  This research project dealt with a particular industry that was monitored for a month giving hourly observations of Nitrogen Oxide concentrations from a stack.  These hourly observations can be changed into many different useful variables.  There are two types of averages that were used.  The first being rolling averages that use three, eight, and twenty-four hour periods to come up with an average every hour.  Second are block averages that happen in three, eight, and twenty-four hour blocks.  For example, a three hour block average is only computed for 8 three hour periods in a day.  Daily maxes are also valuable for controlling certain types of pollution.  All of these different indicators were examined.  Different distributions were constructed to characterize each indicator.  There are known parametric distributions and also nonparametric distributions.   The goodness of fit of each of these known distributions can be tested with the help of some statistical software.  These fitted distributions should characterize the whole population of concentrations instead of just our sample.  From this, we can obtain the standard from the higher percentiles of the fitted distribution, such as the 95th or 99th percentiles.  These provide us with a limit that we know should not be exceeded more than a certain percentage of the time.  Using this limit, the monitoring agency has a method of deciding when an industry is exceeding the acceptable amount of pollution.

 
 
 
 
 
 
 
 
 



 
 
 
 
 
 
 
 
Student Author(s): 

Kadima, Norbert T.

Home Institution:
North Carolina State University

Program:

Statistics VIGRE Traineeship

Department(s):

Statistics

Research Mentor(s)

Daowen Zhang/Statistics

Title of Presentation:

Selecting the Most Predictable Variable of Bone Mineral Density from a Collection of Measurements

 

 

The best way to determine bone density is to have a bone mass measurement (called bone mineral density or BMD test).  Bone mineral density measurements are obtained annually in many sites throughout the country.  These measures provide an indication of bone strength and predisposition to sustain fracture.  BMD in hip is examined in this longitudinal study of 636 women taken from Swan (Study of Women’s health Across the Nation) data and a collection of seven variables was recorded.  The different variables are number of days of collection (totdays), creatinine adjusted progestrone (pdgadj), bone mineral density (bmd_tot), day, body mass index (bmi), cohid, and age.After analysis using a regression model with Swan data, I found that body mass index is the most predictive of BMD.

 

 
 
 
 
 
 
 



 
 
 
 
 
 
 
 

Student Author(s): 

Kalendra, Eric J.

Home Institution:

North Carolina State University

Program:

Statistics VIGRE Traineeship

Department(s):

Statistics

Research Mentor(s)

Kevin Gross/Statistics

Title of Presentation:

Estimating the Abundance of Endangered Butterfly Species from Time Series of Count Data

 

 

Several butterfly species with discrete generations are rare or endangered.  For these species, a model based on Manly (1974) and Zonneveld (1991) can be used to estimate both total population size and death rate from transect count data.  Here, we study the coverage rates of confidence intervals for estimated population size produced by the Manly-Zonneveld model.  When butterfly population dynamics are deterministic, actual coverage rates of confidence intervals are close to the nominal 95% level.  However, under the more reasonable assumption of stochastic population dynamics, confidence interval coverage rates are unacceptably low.  We propose a parametric bootstrap as an alternative for generating confidence intervals, and show that the resulting coverage rates are much closer to the nominal 95%.   This procedure improves the Manly-Zonneveld by quantifying the uncertainty in estimated butterfly abundance more accurately.

 

 


 
 
Student Author(s): 
Pitts, Cathy
Bowes, Joshua
Keene, Nathan
Home Institution:
North Carolina State University
Program:
Independent Researcher
Department(s):

Biological Sciences

Computer Science
Research Mentor(s)

William F. Hunt, Jr./Statistics

Mark Sherriff/Computer Science
Title of Presentation:
Using the EPA Toxic Release Inventory to Compare North Carolina County 
Styrene Releases 
 

                                                           

EPA created the Toxic Release Inventory (TRI) in response to the “Emergency Planning and Community Right-To-Know Act” of 1986.  The public can access the TRI Explorer on EPA’s website.  The TRI Explore provides citizens with the means to identify industry reported routine releases of over 650 chemical materials.  The database spans the years 1988 to 2002 and includes more than 20 industry categories.  The intent of the TRI Explorer is to facilitate communication between the government, the public and industry so that all parties can work together to identify potential problems, set realistic goals and evaluated progress.  One of the chemicals that TRI database contains is styrene.  Our objective is to use the TRI data to create a computer program that compares North Carolina county styrene release trends.  We hope that our research will assist TRI in its effort to make available useful information to TRI Explorer users on chemical releases into communities.  

 



 

 

 [ 2005 Undergraduate Summer Research Symposium Main Page ]


Policy Disclaimer

Last modified July 2005 by Sharon E. Hunt, WordHunting