Skip to Main Content
 

Global Search Box

 
 
 
 

Files

File List


ucin1202404654.pdf (3.6 MB)

ETD Abstract Container

Abstract Header

The Multivariate Generalized Linear Mixed Model for a Joint Modeling Approach for Analysis of Tumor Multiplicity Data: Development and Comparison of Methods

SALISBURY, SHEILIA
http://rave.ohiolink.edu/etdc/view?acc_num=ucin1202404654

Abstract Details

2008, PhD, University of Cincinnati, Medicine : Biostatistics (Environmental Health).
A Monte Carlo simulation was used to examine the small-sample properties of GLIMMIX when two correlated outcome variables, cancer and dysplasia, are modeled jointly. The data were simulated using a range of correlations, variances, samples sizes, treatment group differences, and a negative binomial distribution for each outcome. The simulated data were analyzed using three statistical methods. Correlations were accounted for by modeling both outcomes jointly using GLIMMIX. Separate analyses were done for each outcome using the generalized linear model (GENMOD) and the Non-parametric Kruskal-Wallis test. Efficiency, bias, standard deviation of bias, and goodness of fit for GLIMMIX were compared to GENMOD and efficiency was compared to the Kruskal-Wallis test. When the sample size for each group was 56 or more, the shape of the distributions were similar and the variances defined as a multiple of the mean were small, GLIMMIX was more efficient than GENMOD or Kruskal-Wallis. Goodness of fit statistics were close to one for GLIMMIX and greater than one for GENMOD. With larger variances, Kruskal-Wallis was more efficient. When the distributions of the two groups were slightly different, GENMOD was more efficient with small variances and group samples of 56 or more. None of the methods were appropriate with large variances. Choice of appropriate analysis method for addressing correlated outcomes should be based on underlying distributions, sample size, and variance.
Dr. C. Buncher (Advisor)
Count Data; Correlated; Negative Binomial; Monte Carlo; GLIMMIX; GENMOD; Lognormal-Poisson; Cancer; Simulation

Recommended Citations

Citations

  • SALISBURY, S. (2008). The Multivariate Generalized Linear Mixed Model for a Joint Modeling Approach for Analysis of Tumor Multiplicity Data: Development and Comparison of Methods [Doctoral dissertation, University of Cincinnati]. OhioLINK Electronic Theses and Dissertations Center. http://rave.ohiolink.edu/etdc/view?acc_num=ucin1202404654

    APA Style (7th edition)

  • SALISBURY, SHEILIA. The Multivariate Generalized Linear Mixed Model for a Joint Modeling Approach for Analysis of Tumor Multiplicity Data: Development and Comparison of Methods. 2008. University of Cincinnati, Doctoral dissertation. OhioLINK Electronic Theses and Dissertations Center, http://rave.ohiolink.edu/etdc/view?acc_num=ucin1202404654.

    MLA Style (8th edition)

  • SALISBURY, SHEILIA. "The Multivariate Generalized Linear Mixed Model for a Joint Modeling Approach for Analysis of Tumor Multiplicity Data: Development and Comparison of Methods." Doctoral dissertation, University of Cincinnati, 2008. http://rave.ohiolink.edu/etdc/view?acc_num=ucin1202404654

    Chicago Manual of Style (17th edition)

ucin1202404654
594
© 2008, all rights reserved.
This open access ETD is published by University of Cincinnati and OhioLINK.