Link to the University of Pittsburgh Homepage
Link to the University Library System Homepage Link to the Contact Us Form

Rationale for Choosing an Explicit Correlation Structure in a Multilevel Analysis with Bivariate Outcome.

Atem, Folefac Desire' (2010) Rationale for Choosing an Explicit Correlation Structure in a Multilevel Analysis with Bivariate Outcome. Doctoral Dissertation, University of Pittsburgh. (Unpublished)

Primary Text

Download (739kB) | Preview


The analysis of multileveled data with bivariate outcomes is very common in the fields of education, health economics and health service research. Modeling bivariate outcomes is very useful in HIV research where the joint evolution of HIV RNA and CD4+t lymphocytes in a cohort of HIV-1 infected patient treated with active antiretroviral treatment. The use of the MIXED model method and the Generalized Estimating Equations (GEE) are the most influential recent developments in statistical practice analysis techniques used in analyzing such data. The linear mixed model takes into account all available information and accounts for both serial and cross correlation. The efficiency of the model depends on the correlation structure. Our simulations studies reveal that for smaller clusters the independent and the unstructured are highly favored while for larger clusters the independent models yields estimates with the least standard errors. Additionally, we looked at cases where the data is clustered but not longitudinal. In these cases, the compound symmetry model performed best. Furthermore, our results show that in some cases, the unstructured correlation model tend to have the smallest AICC and BIC but its estimates do not always produce estimates with the smallest standard errors. In this dissertation we formulated a rationale in choosing an explicit working correlation structures for modeling multilevel data with bivariate outcomes. We also simulated different types of data with bivariate outcomes with missingness. To guide our strategy the model selection strategies were based on optimizing AIC, CAIC, AICC BIC and standard error of estimates .Our model has particular public health importance in clinical trials where the clinician may be interested in the joint evolution HIV RNA and CD4+t lymphocytes in a cohort of HIV-1 infected patients treated with active antiretroviral drugs.


Social Networking:
Share |


Item Type: University of Pittsburgh ETD
Status: Unpublished
CreatorsEmailPitt UsernameORCID
Atem, Folefac Desire'
ETD Committee:
TitleMemberEmail AddressPitt UsernameORCID
Committee ChairAnderson, Stewartsja@pitt.eduSJA
Committee MemberWahed, Abduswahed@pitt.eduWAHED
Committee MemberRockette, Howard Eherbst@pitt.eduHERBST
Committee MemberSharma, Ravirks1946@pitt.eduRKS1946
Date: 28 September 2010
Date Type: Completion
Defense Date: 28 August 2010
Approval Date: 28 September 2010
Submission Date: 6 August 2010
Access Restriction: 5 year -- Restrict access to University of Pittsburgh for a period of 5 years.
Institution: University of Pittsburgh
Schools and Programs: School of Public Health > Biostatistics
Degree: PhD - Doctor of Philosophy
Thesis Type: Doctoral Dissertation
Refereed: Yes
Uncontrolled Keywords: Model selection; Mixed model; Random co efficient
Other ID:, etd-08062010-131626
Date Deposited: 10 Nov 2011 19:57
Last Modified: 15 Nov 2016 13:48


Monthly Views for the past 3 years

Plum Analytics

Actions (login required)

View Item View Item