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

An Imputation method under a pseudolikelihood method for analysis of multivariate missing data

Kwon, Yu Mi (2010) An Imputation method under a pseudolikelihood method for analysis of multivariate missing data. Doctoral Dissertation, University of Pittsburgh. (Unpublished)

Primary Text

Download (389kB) | Preview


Missing data are prevalent in many public health studies for various reasons. For example, some subjects do not answer certain questions in a survey, or some subjects drop out of a longitudinal study prematurely. It is important to develop statistical methodologies to appropriately address missing data in order to reach valid conclusions. For regression analysis on data with missing values in the response variable, when data are not missing at random, usually the missing-data mechanism needs to be modeled. When the missingness only depends on the response variable, a pseudolikelihoodmethod that avoids modeling the nonignorable missing-data mechanism was developed in the past. A corresponding mean imputation method was used to impute the missing responses under this pseudolikelihood method. In this dissertation, we consider the inference on the moments of the response variable for missing data analyzed by this pseudolikelihood method. At first, we compared three methods: the delta method, the bootstrap method and a re-sampling method, for estimating the variance of the corresponding pseudolikelihood estimate in simulation studies. Second, we modified that mean imputation method and developed a corresponding stochastic imputation method. Multiple imputations were subsequently used to obtain estimates of the moments and the corresponding variance estimates. We compared the performance of these two imputation methods in simulation studies and illustrated them through analysis of the data from a Schizophrenia clinical trial. Compared to the mean imputation method, the stochastic imputation method leads to less and negligible bias.


Social Networking:
Share |


Item Type: University of Pittsburgh ETD
Status: Unpublished
CreatorsEmailPitt UsernameORCID
Kwon, Yu
ETD Committee:
TitleMemberEmail AddressPitt UsernameORCID
Committee ChairTang, Gonggot1@pitt.eduGOT1
Committee MemberWahed, Abduswaheda@edc.pitt.eduWAHED
Committee MemberYouk, Adaayouk@pitt.eduAYOUK
Committee MemberRen, Dianxudir8@pitt.eduDIR8
Date: 29 September 2010
Date Type: Completion
Defense Date: 10 August 2010
Approval Date: 29 September 2010
Submission Date: 24 July 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: Missing- Data Mechanism; Multivariate; NMAR; Pseudolikelihood; Imputation; Variance
Other ID:, etd-07242010-185029
Date Deposited: 10 Nov 2011 19:53
Last Modified: 15 Nov 2016 13:46


Monthly Views for the past 3 years

Plum Analytics

Actions (login required)

View Item View Item