Marron, Megan
(2015)
Cross-validation in group-based latent trajectory modeling when assuming a censored normal model.
Master's Thesis, University of Pittsburgh.
(Unpublished)
Abstract
Group-based latent trajectory modeling (GBLTM) is a relatively recent addition to methodology for analyzing repeated measures of a variable over time. It is implemented using SAS procedure TRAJ for a zero-inflated Poisson (ZIP) model, censored normal (CNORM) model, and logistic model. Cross-validation (CV) in GBLTM is used as an alternative tool to the Bayesian Information Criterion (BIC) for determining the optimal number of distinct latent subgroups in a sample. CV in GBLTM when assuming a ZIP model is implemented using the crimCV package in R. In this thesis, the use of CV in GBLTM is furthered by applying it when assuming a CNORM model and examining the consistency of results when considering multiple types of CV. This method is applied to a Hepatitis C (HCV) study to determine whether patterns of depressive symptoms in HCV patients treated with interferon-based therapy form clinically meaningful distinct subgroups. When applied to the HCV study, CV was a conservative approach to model selection compared with BIC; CV suggested a two-group model, whereas BIC suggested a five-group model. However, when visually examining the data, a three-group model appeared to capture the heterogeneity in the HCV sample best. Therefore, BIC and CV should not be used alone to determine the optimal number of distinct latent subgroups in a sample, but rather used to make an educated judgment on the number of subgroups that describes the heterogeneity best. Whether or not CV is truly a conservative approach to model selection compared with BIC is still unknown, CV and BIC should be further explored using other datasets and simulations. The public health significance of this thesis is exploring statistical tools used for determining the optimal number of distinct latent subgroups in GBLTM, where knowledge of the factors that predispose individuals to less favorable trajectory groups, can lead to targeted preventions.
Share
| Citation/Export: |
|
| Social Networking: |
|
Details
| Item Type: |
University of Pittsburgh ETD
|
| Status: |
Unpublished |
| Creators/Authors: |
|
| ETD Committee: |
|
| Date: |
28 January 2015 |
| Date Type: |
Publication |
| Defense Date: |
5 December 2014 |
| Approval Date: |
28 January 2015 |
| Submission Date: |
9 December 2014 |
| Access Restriction: |
1 year -- Restrict access to University of Pittsburgh for a period of 1 year. |
| Number of Pages: |
74 |
| Institution: |
University of Pittsburgh |
| Schools and Programs: |
School of Public Health > Biostatistics |
| Degree: |
MS - Master of Science |
| Thesis Type: |
Master's Thesis |
| Refereed: |
Yes |
| Uncontrolled Keywords: |
cross-validation, group-based latent trajectory modeling, proc TRAJ, Bayesian information criterion, hepatitis C, depression, sleep quality |
| Date Deposited: |
28 Jan 2015 15:45 |
| Last Modified: |
19 Dec 2016 14:42 |
| URI: |
http://d-scholarship.pitt.edu/id/eprint/23833 |
Metrics
Monthly Views for the past 3 years
Plum Analytics
Actions (login required)
 |
View Item |
![[img]](http://d-scholarship.pitt.edu/style/images/fileicons/application_pdf.png)
![[feed]](/images/feed-icon-32x32.png){kind=icon}