The analysis of repeated measures data is important for most studies in the Division of Epidemiology, Statistics, and Prevention Research (DESPR) and in the NICHD as a whole. There are a number of areas in which additional methodological research is needed in order to appropriate analyze data from research studies. Missing data is a common problem when analyzing data from longitudinal studies. Investigators need to account for this missing data when analyzing data from their studies. The development of new methods for analyzing continuous or discrete repeated measures data is important area of methodological research. Although there is a wide literature for a single longitudinal measurement, there has been little work done on appropriately accounting for missing data when the number of longitudinal measurements is large. We will development new methods for addressing this problem under this project. Developing approach for jointly modeling time-to-event and repeated measures is currently an active area of biostatistical research. Most of this work focuses on modeling a single repeated biomarker and the time to an event. Jointly modeling multiple repeated biomarkers and time to event data is an important, yet difficult problem. In this project, we will investigate this problem with both frequentist and Bayesian approaches. Important studies that will illustrate this methodology include the Biocycle study in the Epidemiology Branch and the Natural Driving Study in the Prevention Branch. Estimating the patterns in key biomarkers over the menstrual period is important for understanding the biology of the menstrual period. Also, important is understanding how environmental factors affect the relationship between these biomarkers over the cycle. The Bio-cycle study was designed to address these scientific issues. Using the Bio-cycle data as a motivating example, we will develop new approaches for analyzing such data. Understanding ranges of normal growth is an important research area within DESPR and in the institute as a whole. For example, studying ranges of normal fetal growth as measured by ultra-sound is important in identifying fetuses who are growing abnormally. For many of these ultra-sound longitudinal fetal studies, the number of measurements may be related to the underlying growth curve of the particular fetus since continued follow-up may be based on a clinical decision (i.e., more ultra-sound measurements may be taken on fetuses who grow slowly). This type of observation mechanism may result in informative number of measurements which needs to be taken into account in the statistical modeling. We will develop new methods for non-linear growth models and apply them to data from our fetal growth studies

Project Start
Project End
Budget Start
Budget End
Support Year
4
Fiscal Year
2012
Total Cost
$179,407
Indirect Cost
City
State
Country
Zip Code
Lai, Yinglei; Albert, Paul S (2014) Identifying multiple change points in a linear mixed effects model. Stat Med 33:1015-28
Winer, Karen K; Fulton, Kara A; Albert, Paul S et al. (2014) Effects of pump versus twice-daily injection delivery of synthetic parathyroid hormone 1-34 in children with severe congenital hypoparathyroidism. J Pediatr 165:556-63.e1
Cheon, Kyeongmi; Thoma, Marie E; Kong, Xiangrong et al. (2014) A mixture of transition models for heterogeneous longitudinal ordinal data: with applications to longitudinal bacterial vaginosis data. Stat Med 33:3204-13
Liu, Danping; Albert, Paul S (2014) Combination of longitudinal biomarkers in predicting binary events. Biostatistics 15:706-18
McLain, Alexander C; Albert, Paul S (2014) Modeling longitudinal data with a random change point and no time-zero: applications to inference and prediction of the labor curve. Biometrics 70:1052-60
Ogbagaber, Semhar; Albert, Paul S; Lewin, Daniel et al. (2012) Summer activity patterns among teenage girls: harmonic shape invariant modeling to estimate circadian cycles. J Circadian Rhythms 10:2
Cheon, Kyeongmi; Albert, Paul S; Zhang, Zhiwei (2012) The impact of random-effect misspecification on percentile estimation for longitudinal growth data. Stat Med 31:3708-18
Zhang, Bo; Chen, Zhen; Albert, Paul S (2012) Latent class models for joint analysis of disease prevalence and high-dimensional semicontinuous biomarker data. Biostatistics 13:74-88
Fan, Ruzong; Albert, Paul S; Schisterman, Enrique F (2012) A discussion of gene-gene and gene-environment interactions and longitudinal genetic analysis of complex traits. Stat Med 31:2565-8
Albert, Paul S (2012) A linear mixed model for predicting a binary event from longitudinal data under random effects misspecification. Stat Med 31:143-54

Showing the most recent 10 out of 26 publications