D1.

Primary Faculty co-Advisors

Dr. Eugene Kennedy, ELRC (Educational Research Methology)

Dr. Charles J. Monlezun, Experimental Statistics (Mathematical Statistics)

Off-campus Participant: Gene Fields (Louisiana Board of Regents)

Technical Proposal:

Survival analysis, also known as time to event data analysis or time-till-failure analysis, is the name given to a class of statistical models where time is the unknown random variable of interest. There are three things needed for survival analysis. First the event of interest most have a well-defined beginning, second the scale or method for measuring time must be known and finally the event must have a well-defined ending.

There are times in survival analysis when the event time is not know, these times are known as censoring. Some common reasons for censoring are the constant need to look at life test data before all units have failed, and when actual response values cannot be observed for some or all units.

Examples of survival analysis are: In economics the time until an unemployed person becomes employed again or the duration of a strike. In sociology the time until a marriage ends in divorce. In engineering time until a light bulb fails. In medicine the time until a person dies from a sickness.

The question of why to use survival analysis is an easy one to answer. Survival analysis is used because censoring and time-dependent covariates are two aspects of survival data that conventional statistical methods have trouble dealing with. The following example illustrates these two concepts: A sample of 432 inmates from the Maryland state prison system were released and followed for one year. The event of interest was the first time a form inmate is arrested. The study was done to determine how the occurrence and timing of arrests depended on several covariates. Some of these covariates (such as race, age at release, and number of previous convictions) remained constant over the one-year study, but other covariates (such as marital status and employment status) could change over time during the study. One conventional method of analysis is to use logit analysis with a dichotomous dependent variable: arrested or not arrested. But logit analysis will ignore the information on the timing of arrests. It is reasonable to assume that an individual who is arrested one week after being released, on average, has a higher propensity than those arrested 52 weeks after being released. By ignoring this information one would think this will reduce the precision of the estimation.

For the purposes of this project survival analysis will be used to determine the retention rate of university students at four-year public colleges in Louisiana. This is a useful tool in helping university officials determine which factors lead to keeping students at their universities and what factors may lead to a student not staying at a university. This will also help Louisiana legislators in determining which schools are making the most of their monetary funds.

Number of IGERT apprentices to be recruited and probable home departments:

Consistency with the Macromolecular Education, Research & Training theme: The project will allow for hands on applications of survival analysis. The project will allow for use of different statistical software and their different applications. .

How does the project form a vector cross-product of existing research themes by the participants?

How do students benefit from the team-oriented research, beyond what would be available to them from either advisor separately?

The team-oriented research will bridge the gap between statistical theory and the application of statistics in the real world. The students will benefit be learning what statistical analysis works in a real world situation and how to deal with the limitation of statistical theory.

Briefly describe the support level available to each individual faculty or off-campus participant (i.e., without IGERT)

Interdisciplinary strengths of the team project: With team members in the fields of statistics, education research, and the governing board of public Louisiana colleges team members will interact, trade ideas and learn what works and what does not work, why something works and why it does not work. All team members will benefit from the cross-product of these three areas. Each member will have a better understanding of how their work is applied in a separate field and used their work is used by some one in another field. Members will have a better understanding of what it done in order to get the information they are using in each field. These three fields are connected and a better understanding of how the other areas function will strength the support between the three areas of research.

Commitment of faculty & off-campus participants to work side-by-side with apprentices:

References: