Combining Information from Multiple Sources in Bayesian Modeling

Surveys can collect important data that inform policy decisions and drive social science research. Large government surveys collect information from the U.S. population on a wide range of topics, including demographics, education, employment, and lifestyle. Analysis of survey data presents unique challenges. In particular, one needs to account for missing data, for complex sampling designs, and for measurement error. Conceptually, a survey organization could spend lots of resources getting high-quality responses from a simple random sample, resulting in survey data that are easy to analyze. However, this scenario often is not realistic. To address these practical issues, survey organizations can leverage the information available from other sources of data. For example, in longitudinal studies that suffer from attrition, they can use the information from refreshment samples to correct for potential attrition bias. They can use information from known marginal distributions or survey design to improve inferences. They can use information from gold standard sources to correct for measurement error.

This thesis presents novel approaches to combining information from multiple sources that address the three problems described above.

The first method addresses nonignorable unit nonresponse and attrition in a panel survey with a refreshment sample. Panel surveys typically suffer from attrition, which can lead to biased inference when basing analysis only on cases that complete all waves of the panel. Unfortunately, the panel data alone cannot inform the extent of the bias due to attrition, so analysts must make strong and untestable assumptions about the missing data mechanism. Many panel studies also include refreshment samples, which are data collected from a random sample of new

individuals during some later wave of the panel. Refreshment samples offer information that can be utilized to correct for biases induced by nonignorable attrition while reducing reliance on strong assumptions about the attrition process. To date, these bias correction methods have not dealt with two key practical issues in panel studies: unit nonresponse in the initial wave of the panel and in the

refreshment sample itself. As we illustrate, nonignorable unit nonresponse

can significantly compromise the analyst's ability to use the refreshment samples for attrition bias correction. Thus, it is crucial for analysts to assess how sensitive their inferences---corrected for panel attrition---are to different assumptions about the nature of the unit nonresponse. We present an approach that facilitates such sensitivity analyses, both for suspected nonignorable unit nonresponse

in the initial wave and in the refreshment sample. We illustrate the approach using simulation studies and an analysis of data from the 2007-2008 Associated Press/Yahoo News election panel study.

The second method incorporates informative prior beliefs about

marginal probabilities into Bayesian latent class models for categorical data.

The basic idea is to append synthetic observations to the original data such that

(i) the empirical distributions of the desired margins match those of the prior beliefs, and (ii) the values of the remaining variables are left missing. The degree of prior uncertainty is controlled by the number of augmented records. Posterior inferences can be obtained via typical MCMC algorithms for latent class models, tailored to deal efficiently with the missing values in the concatenated data.

We illustrate the approach using a variety of simulations based on data from the American Community Survey, including an example of how augmented records can be used to fit latent class models to data from stratified samples.

The third method leverages the information from a gold standard survey to model reporting error. Survey data are subject to reporting error when respondents misunderstand the question or accidentally select the wrong response. Sometimes survey respondents knowingly select the wrong response, for example, by reporting a higher level of education than they actually have attained. We present an approach that allows an analyst to model reporting error by incorporating information from a gold standard survey. The analyst can specify various reporting error models and assess how sensitive their conclusions are to different assumptions about the reporting error process. We illustrate the approach using simulations based on data from the 1993 National Survey of College Graduates. We use the method to impute error-corrected educational attainments in the 2010 American Community Survey using the 2010 National Survey of College Graduates as the gold standard survey.

Harmonic source wavefront aberration correction for ultrasound imaging.

A method is proposed which uses a lower-frequency transmit to create a known harmonic acoustical source in tissue suitable for wavefront correction without a priori assumptions of the target or requiring a transponder. The measurement and imaging steps of this method were implemented on the Duke phased array system with a two-dimensional (2-D) array. The method was tested with multiple electronic aberrators [0.39π to 1.16π radians root-mean-square (rms) at 4.17 MHz] and with a physical aberrator 0.17π radians rms at 4.17 MHz) in a variety of imaging situations. Corrections were quantified in terms of peak beam amplitude compared to the unaberrated case, with restoration between 0.6 and 36.6 dB of peak amplitude with a single correction. Standard phantom images before and after correction were obtained and showed both visible improvement and 14 dB contrast improvement after correction. This method, when combined with previous phase correction methods, may be an important step that leads to improved clinical images.

Plate-specific gain map correction for the improvement of detective quantum efficiency in computed radiography.

PURPOSE: The purpose of this work is to improve the noise power spectrum (NPS), and thus the detective quantum efficiency (DQE), of computed radiography (CR) images by correcting for spatial gain variations specific to individual imaging plates. CR devices have not traditionally employed gain-map corrections, unlike the case with flat-panel detectors, because of the multiplicity of plates used with each reader. The lack of gain-map correction has limited the DQE(f) at higher exposures with CR. This current work describes a feasible solution to generating plate-specific gain maps. METHODS: Ten high-exposure open field images were taken with an RQA5 spectrum, using a sixth generation CR plate suspended in air without a cassette. Image values were converted to exposure, the plates registered using fiducial dots on the plate, the ten images averaged, and then high-pass filtered to remove low frequency contributions from field inhomogeneity. A gain-map was then produced by converting all pixel values in the average into fractions with mean of one. The resultant gain-map of the plate was used to normalize subsequent single images to correct for spatial gain fluctuation. To validate performance, the normalized NPS (NNPS) for all images was calculated both with and without the gain-map correction. Variations in the quality of correction due to exposure levels, beam voltage/spectrum, CR reader used, and registration were investigated. RESULTS: The NNPS with plate-specific gain-map correction showed improvement over the noncorrected case over the range of frequencies from 0.15 to 2.5 mm(-1). At high exposure (40 mR), NNPS was 50%-90% better with gain-map correction than without. A small further improvement in NNPS was seen from carefully registering the gain-map with subsequent images using small fiducial dots, because of slight misregistration during scanning. Further improvement was seen in the NNPS from scaling the gain map about the mean to account for different beam spectra. CONCLUSIONS: This study demonstrates that a simple gain-map can be used to correct for the fixed-pattern noise in a given plate and thus improve the DQE of CR imaging. Such a method could easily be implemented by manufacturers because each plate has a unique bar code and the gain-map for all plates associated with a reader could be stored for future retrieval. These experiments indicated that an improvement in NPS (and hence, DQE) is possible, depending on exposure level, over a wide range of frequencies with this technique.

Bound on quantum computation time: Quantum error correction in a critical environment

We obtain an upper bound on the time available for quantum computation for a given quantum computer and decohering environment with quantum error correction implemented. First, we derive an explicit quantum evolution operator for the logical qubits and show that it has the same form as that for the physical qubits but with a reduced coupling strength to the environment. Using this evolution operator, we find the trace distance between the real and ideal states of the logical qubits in two cases. For a super-Ohmic bath, the trace distance saturates, while for Ohmic or sub-Ohmic baths, there is a finite time before the trace distance exceeds a value set by the user. © 2010 The American Physical Society.

3D refraction correction and extraction of clinical parameters from spectral domain optical coherence tomography of the cornea.

Capable of three-dimensional imaging of the cornea with micrometer-scale resolution, spectral domain-optical coherence tomography (SDOCT) offers potential advantages over Placido ring and Scheimpflug photography based systems for accurate extraction of quantitative keratometric parameters. In this work, an SDOCT scanning protocol and motion correction algorithm were implemented to minimize the effects of patient motion during data acquisition. Procedures are described for correction of image data artifacts resulting from 3D refraction of SDOCT light in the cornea and from non-idealities of the scanning system geometry performed as a pre-requisite for accurate parameter extraction. Zernike polynomial 3D reconstruction and a recursive half searching algorithm (RHSA) were implemented to extract clinical keratometric parameters including anterior and posterior radii of curvature, central cornea optical power, central corneal thickness, and thickness maps of the cornea. Accuracy and repeatability of the extracted parameters obtained using a commercial 859nm SDOCT retinal imaging system with a corneal adapter were assessed using a rigid gas permeable (RGP) contact lens as a phantom target. Extraction of these parameters was performed in vivo in 3 patients and compared to commercial Placido topography and Scheimpflug photography systems. The repeatability of SDOCT central corneal power measured in vivo was 0.18 Diopters, and the difference observed between the systems averaged 0.1 Diopters between SDOCT and Scheimpflug photography, and 0.6 Diopters between SDOCT and Placido topography.

Evaluation of Altered Kras Codon Bias and NOS Inhibition During Lung Tumorigenesis

The small GTPases HRAS, NRAS and KRAS are mutated in approximately one-third of all human cancers, rendering the proteins constitutively active and oncogenic. Lung cancer is the leading cause of cancer deaths worldwide, and more than 20% of human lung cancers harbor mutations in RAS, with 98% of those occurring in the KRAS isoform. While there have been many advances in the understanding of KRAS–driven lung tumorigenesis, it remains a therapeutic challenge. To further this understanding and assess novel approaches for treatment, I have investigated two aspects of Kras–driven tumorigenesis in the lung:

(I) Despite nearly identical protein sequences, the three RAS proto-oncogenes exhibit divergent codon usage. Of the three isoforms, KRAS contains the most rare codons resulting in lower levels of KRAS protein expression relative to HRAS and NRAS. To determine the consequences of rare codon bias during de novo tumorigenesis, we created a knock-in Krasex3op mouse in which synonymous mutations in exon 3 converted codons from rare to common. These mice had reduced tumor burden and fewer oncogenic mutations in the Krasex3op allele following carcinogen exposure. The reduction in tumorigenesis appeared to be a product of rare codons affecting both the oncogenic and non–oncogenic alleles. Converting rare codons to common codons yielded a more potent oncogenic allele that promoted growth arrest and enhanced tumor suppression by the non-oncogenic allele. Thus, rare codons play an integral role in Kras tumorigenesis.

(II) Lung cancer patients exhale higher levels of NO and iNOS-/- mice are resistant to chemically induced lung tumorigenesis. I hypothesize that NO promotes Kras–driven lung adenocarcinoma, and NOS inhibition may decrease Kras–driven lung tumorigenesis. To test this hypothesis, I assessed efficacy of the NOS inhibitor L–NAME in a genetically engineered mouse model of Kras-driven lung adenocarcinoma. Adenoviral Cre recombinase was delivered into the lungs intranasally, resulting in expression of oncogenic KrasG12D and dominant-negative Trp53R172H in lung epithelial cells. L–NAME treatment was provided in the water and continued until survival endpoints. In this model, L–NAME treatment decreased tumor growth and prolonged survival. These data establish a potential clinical role for NOS inhibition in lung cancer treatment.

Hybrid error correction and de novo assembly of single-molecule sequencing reads.

Single-molecule sequencing instruments can generate multikilobase sequences with the potential to greatly improve genome and transcriptome assembly. However, the error rates of single-molecule reads are high, which has limited their use thus far to resequencing bacteria. To address this limitation, we introduce a correction algorithm and assembly strategy that uses short, high-fidelity sequences to correct the error in single-molecule sequences. We demonstrate the utility of this approach on reads generated by a PacBio RS instrument from phage, prokaryotic and eukaryotic whole genomes, including the previously unsequenced genome of the parrot Melopsittacus undulatus, as well as for RNA-Seq reads of the corn (Zea mays) transcriptome. Our long-read correction achieves >99.9% base-call accuracy, leading to substantially better assemblies than current sequencing strategies: in the best example, the median contig size was quintupled relative to high-coverage, second-generation assemblies. Greater gains are predicted if read lengths continue to increase, including the prospect of single-contig bacterial chromosome assembly.

Barriers to and facilitators of interventions to counter publication bias: thematic analysis of scholarly articles and stakeholder interviews.

BACKGROUND: When the nature and direction of research results affect their chances of publication, a distortion of the evidence base - termed publication bias - results. Despite considerable recent efforts to implement measures to reduce the non-publication of trials, publication bias is still a major problem in medical research. The objective of our study was to identify barriers to and facilitators of interventions to prevent or reduce publication bias. METHODS: We systematically reviewed the scholarly literature and extracted data from articles. Further, we performed semi-structured interviews with stakeholders. We performed an inductive thematic analysis to identify barriers to and facilitators of interventions to counter publication bias. RESULTS: The systematic review identified 39 articles. Thirty-four of 89 invited interview partners agreed to be interviewed. We clustered interventions into four categories: prospective trial registration, incentives for reporting in peer-reviewed journals or research reports, public availability of individual patient-level data, and peer-review/editorial processes. Barriers we identified included economic and personal interests, lack of financial resources for a global comprehensive trial registry, and different legal systems. Facilitators identified included: raising awareness of the effects of publication bias, providing incentives to make data publically available, and implementing laws to enforce prospective registration and reporting of clinical trial results. CONCLUSIONS: Publication bias is a complex problem that reflects the complex system in which it occurs. The cooperation amongst stakeholders to increase public awareness of the problem, better tailoring of incentives to publish, and ultimately legislative regulations have the greatest potential for reducing publication bias.

Production and recognition bias of stylistic sentences using a story reading task.

Four experiments examined participants' ability to produce surface characteristics of sentences using an on-line story reading task. Participants read a series of stories in which either all, or the majority of sentences were written in the same "style," or surface form. Twice per story, participants were asked to fill in a blank consistent with the story. For sentences that contained three stylistic regularities, participants imitated either all three characteristics (Experiment 2) or two of the three characteristics (Experiment 1), depending on the proportion of in-style sentences. Participants demonstrated a recognition bias for the read style in an unannounced recognition task. When participants read stories in which the two styles were the dative/double object alternation, participants demonstrated a syntactic priming effect in the cloze task, but no consistent recognition bias in a later recognition test (Experiments 3 and 4).

Hostile attributional bias and aggressive behavior in global context.

We tested a model that children's tendency to attribute hostile intent to others in response to provocation is a key psychological process that statistically accounts for individual differences in reactive aggressive behavior and that this mechanism contributes to global group differences in children's chronic aggressive behavior problems. Participants were 1,299 children (mean age at year 1 = 8.3 y; 51% girls) from 12 diverse ecological-context groups in nine countries worldwide, followed across 4 y. In year 3, each child was presented with each of 10 hypothetical vignettes depicting an ambiguous provocation toward the child and was asked to attribute the likely intent of the provocateur (coded as benign or hostile) and to predict his or her own behavioral response (coded as nonaggression or reactive aggression). Mothers and children independently rated the child's chronic aggressive behavior problems in years 2, 3, and 4. In every ecological group, in those situations in which a child attributed hostile intent to a peer, that child was more likely to report that he or she would respond with reactive aggression than in situations when that same child attributed benign intent. Across children, hostile attributional bias scores predicted higher mother- and child-rated chronic aggressive behavior problems, even controlling for prior aggression. Ecological group differences in the tendency for children to attribute hostile intent statistically accounted for a significant portion of group differences in chronic aggressive behavior problems. The findings suggest a psychological mechanism for group differences in aggressive behavior and point to potential interventions to reduce aggressive behavior.

Explicit DNase sequence bias modeling enables high-resolution transcription factor footprint detection.

DNaseI footprinting is an established assay for identifying transcription factor (TF)-DNA interactions with single base pair resolution. High-throughput DNase-seq assays have recently been used to detect in vivo DNase footprints across the genome. Multiple computational approaches have been developed to identify DNase-seq footprints as predictors of TF binding. However, recent studies have pointed to a substantial cleavage bias of DNase and its negative impact on predictive performance of footprinting. To assess the potential for using DNase-seq to identify individual binding sites, we performed DNase-seq on deproteinized genomic DNA and determined sequence cleavage bias. This allowed us to build bias corrected and TF-specific footprint models. The predictive performance of these models demonstrated that predicted footprints corresponded to high-confidence TF-DNA interactions. DNase-seq footprints were absent under a fraction of ChIP-seq peaks, which we show to be indicative of weaker binding, indirect TF-DNA interactions or possible ChIP artifacts. The modeling approach was also able to detect variation in the consensus motifs that TFs bind to. Finally, cell type specific footprints were detected within DNase hypersensitive sites that are present in multiple cell types, further supporting that footprints can identify changes in TF binding that are not detectable using other strategies.

Associations between BMI and home, school and route environmental exposures estimated using GPS and GIS: do we see evidence of selective daily mobility bias in children?

BACKGROUND: This study examined whether objective measures of food, physical activity and built environment exposures, in home and non-home settings, contribute to children's body weight. Further, comparing GPS and GIS measures of environmental exposures along routes to and from school, we tested for evidence of selective daily mobility bias when using GPS data. METHODS: This study is a cross-sectional analysis, using objective assessments of body weight in relation to multiple environmental exposures. Data presented are from a sample of 94 school-aged children, aged 5-11 years. Children's heights and weights were measured by trained researchers, and used to calculate BMI z-scores. Participants wore a GPS device for one full week. Environmental exposures were estimated within home and school neighbourhoods, and along GIS (modelled) and GPS (actual) routes from home to school. We directly compared associations between BMI and GIS-modelled versus GPS-derived environmental exposures. The study was conducted in Mebane and Mount Airy, North Carolina, USA, in 2011. RESULTS: In adjusted regression models, greater school walkability was associated with significantly lower mean BMI. Greater home walkability was associated with increased BMI, as was greater school access to green space. Adjusted associations between BMI and route exposure characteristics were null. The use of GPS-actual route exposures did not appear to confound associations between environmental exposures and BMI in this sample. CONCLUSIONS: This study found few associations between environmental exposures in home, school and commuting domains and body weight in children. However, walkability of the school neighbourhood may be important. Of the other significant associations observed, some were in unexpected directions. Importantly, we found no evidence of selective daily mobility bias in this sample, although our study design is in need of replication in a free-living adult sample.

Correction for Eddy Current-Induced Echo-Shifting Effect in Partial-Fourier Diffusion Tensor Imaging.

In most diffusion tensor imaging (DTI) studies, images are acquired with either a partial-Fourier or a parallel partial-Fourier echo-planar imaging (EPI) sequence, in order to shorten the echo time and increase the signal-to-noise ratio (SNR). However, eddy currents induced by the diffusion-sensitizing gradients can often lead to a shift of the echo in k-space, resulting in three distinct types of artifacts in partial-Fourier DTI. Here, we present an improved DTI acquisition and reconstruction scheme, capable of generating high-quality and high-SNR DTI data without eddy current-induced artifacts. This new scheme consists of three components, respectively, addressing the three distinct types of artifacts. First, a k-space energy-anchored DTI sequence is designed to recover eddy current-induced signal loss (i.e., Type 1 artifact). Second, a multischeme partial-Fourier reconstruction is used to eliminate artificial signal elevation (i.e., Type 2 artifact) associated with the conventional partial-Fourier reconstruction. Third, a signal intensity correction is applied to remove artificial signal modulations due to eddy current-induced erroneous T2(∗) -weighting (i.e., Type 3 artifact). These systematic improvements will greatly increase the consistency and accuracy of DTI measurements, expanding the utility of DTI in translational applications where quantitative robustness is much needed.