There is a class of complex problems that may be too complicated to solve by any single analytical technique. Such problems involve so many measurements of interconnected factors that analysis with a single dimension technique may improve one aspect of the problem while overall achieving little or no improvement. This research examines the utility of modeling a complex problem with multiple statistical techniques integrated to analyze different types of data. The goal was to determine if this integrated approach was feasible and provided significantly better results than a single statistical technique. An application in engineering education was chosen because of the availability and comprehensiveness of the NELS:88 longitudinal dataset. This dataset provided a huge number of variables and 12,144 records of actual students progressing from 8th grade to their final educational outcomes 12 years later.
The probability of earning a Science, Technology, Engineering, or Mathematics (STEM) degree is modeled using variables available in the 8th grade as well as standardized test scores. The variables include demographic, academic performance, and experiential measures. Extensive manipulation of the NELS:88 dataset was conducted to identify the student outcomes, prepare the set of covariates for modeling, and determine when the students final outcome status occurred. The integrated models combined logistic regression, survival analysis, and Receiver Operating Characteristics (ROC) Curve analysis to predict obtaining a STEM degree vs. other outcomes. The results of the integrated models were compared to actual outcomes and the results of separate logistic regression models. Both sets of models provided good predictive accuracy. The feasibility of integrated models for complex problems was confirmed. The integrated approach provided less variability in incorrect STEM predictions, but the improvement was not statistically significant.
The main contribution of this research is designing the integrated model approach and confirming its feasibility. Additional contributions include designing a process to create large multivariate logistic regression models; developing methods for extensive manipulation of a large dataset to adapt it for new analytical purposes; extending the application of logistic regression, survival analysis, and ROC Curve analysis within educational research; and creating a formal definition for STEM that can be statistically verified.
| Identifer | oai:union.ndltd.org:PITT/oai:PITTETD:etd-11212008-141246 |
| Date | 28 January 2009 |
| Creators | Nicholls, Gillian M. |
| Contributors | Dr. Harvey Wolfe, Dr. Henry W. Block Jr., Dr. Kim L. Needy, Dr. Mary Besterfield-Sacre, Dr. Larry J. Shuman |
| Publisher | University of Pittsburgh |
| Source Sets | University of Pittsburgh |
| Language | English |
| Detected Language | English |
| Type | text |
| Format | application/pdf |
| Source | http://etd.library.pitt.edu/ETD/available/etd-11212008-141246/ |
| Rights | unrestricted, I hereby certify that, if appropriate, I have obtained and attached hereto a written permission statement from the owner(s) of each third party copyrighted matter to be included in my thesis, dissertation, or project report, allowing distribution as specified below. I certify that the version I submitted is the same as that approved by my advisory committee. I hereby grant to University of Pittsburgh or its agents the non-exclusive license to archive and make accessible, under the conditions specified below, my thesis, dissertation, or project report in whole or in part in all forms of media, now or hereafter known. I retain all other ownership rights to the copyright of the thesis, dissertation or project report. I also retain the right to use in future works (such as articles or books) all or part of this thesis, dissertation, or project report. |
Page generated in 0.0067 seconds