Proceedings. American Statistical Association. Annual Meeting最新文献

In designing a longitudinal study one needs to decide on two critical components: duration of study and frequency of visits. In addition, other issues involving sample size, power, number of observations per subject must be addressed. If the study is meant to be completed within a certain time frame, would it better to have a fixed time between observations (which might allow the study to terminate early if its objectives are met) or to spread out the observations over the entire study period? At some point during the study, it may be of interest to see if additional data points would contribute substantially. Assume that the longitudinal data will be analyzed using a linear mixed-effects model. In this investigation we use the standard errors of estimates of model parameters as the criterion. We seek to address the issues using three approaches. First, subsets of a data set are constructed in a number of ways and the standard errors are examined. Second, using a variety of designs, the covariance matrix of the fixed-effects is computed and the standard errors are examined. Finally, a simulation study is conducted.

We propose an innovative approach to the problem recently posed by Hall and Schimek (2012): determining at what point the agreement between two rankings of a long list of items degenerates into noise. We modify the method of estimation in Fligner and Verducci's (1988) multistage model for rankings, from maximum likelihood of conditional agreement over a sample of rankings to a locally smooth estimator of agreement. Through simulations we show that this innovation performs very well under several conditions. Some ramifications are discussed as planned extensions.

Cell phone surveys have become increasingly popular and researchers have noted major challenges in conducting cost-effective surveys while achieving high response rates. Previous work has shown that calling strategies that maximize both respondent contact and completed interviews for landline surveys may not be the most cost-effective for cell phone surveys. For example, Montgomery, et al. (2011) found important differences between landline and cell samples for best times to call and declines in contact rates after repeated dialing. Using paradata from the 2010 and 2011 National Flu Surveys (sponsored by the Centers for Disease Control and Prevention), we investigate differences in calling outcomes between landline and cell surveys. Specifically, we predict respondent contact and interview completion using logistic regression models that examine the impact of calling on particular days of the week, certain times of the day, number of previous calls, outcomes of previous calls and length of time between calls. We discuss how these differences can be used to increase the likelihood of contacting cooperative respondents and completing interviews for both sample types.

The investigation of genetic factors that determine differential drug response is a key goal of pharmacogenomics (PGX), and relies on the often-untested assumption that differential response is heritable. While limitations in traditional study design often prohibit heritability (h²) estimates in PGX, new approaches may allow such estimates. We demonstrate an ex vivo model system to determine the h² of drug-induced cell killing and performed genome-wide analysis for gene mapping. The cytotoxic effect of 29 diverse chemotherapeutic agents on lymphoblastoid cell lines (LCLs) derived from family- and population-based cohorts was investigated. We used a high throughput format to determine cytotoxicity of the drugs on LCLs and developed a new evolutionary computation approach to fit response curves for each individual. Variance components analysis determined the h² for each drug response and a wide range of values was observed across drugs. Genome-wide analysis was performed using new analytical approaches. These results lay the groundwork for future studies to uncover genes influencing chemotherapeutic response and demonstrate a new computational framework for performing such analysis.

We developed a multinomial probit model with singular value decomposition for testing a large number of single nucleotide polymorphisms (SNPs) simultaneously, using maximum likelihood estimation and permutation. The method was validated by simulation. We simulated 1000 SNPs, including 9 associated with disease states, and 8 of the 9 were successfully identified. Applying the method to study 32 genes in our Mexican-American samples for association with prediabetes through either impaired glucose tolerance (IGT) or impaired fasting glucose (IFG), we found 3 genes (SORCS1, AMPD1, PPAR) associated with both IGT and IFG, while 5 genes (AMPD2, PRKAA2, C5, TCF7L2, ITR) with the IGT mechanism only and 6 genes (CAPN10, IL4,NOS3, CD14, GCG, SORT1) with the IFG mechanism only. These data suggest that IGT and IFG may indicate different physiological mechanism to prediabetes, via different genetic determinants.