Browsing by Subject "prediction"

Now showing 1 - 5 of 5
  • Thumbnail Image
    Item
    Atrial Fibrillation In Older Adults: Relation To Proteomics, Risk Prediction, And Urban/Rural Disparities In Treatment And Outcomes
    (2020-07) Norby, Faye
    Atrial fibrillation (AF), a cardiac arrhythmia, is a major public health problem. AF is largely a disease of advancing age and contributes to other cardiovascular complications. Identification of novel protein biomarkers could advance understanding of AF mechanisms and may improve the prediction of incident AF. Additionally, it is unknown if disparities exist in AF treatment and outcomes in rural versus urban areas of the US. For manuscripts 1 and 2, we used data from the Atherosclerosis Risk in Communities (ARIC) study, a cohort of older-aged adults in the US. For manuscripts 3 and 4, we used a sample of Medicare beneficiaries enrolled from 2011-2016 with residential zip code categorized into 4 rural/urban areas. In the first manuscript, we examined the association of plasma proteins and identified 40 novel protein biomarkers associated with incident AF. These biomarkers provide insight into mechanistic pathways of AF development. In the second manuscript, we derived and validated a series of 5-year incident AF prediction models that are better targeted and calibrated to older populations. Incorporating biomarkers, including proteomics data, into the models improved AF risk prediction. In the third and fourth manuscripts, we examined the initiation of anticoagulation use and compared the risks of subsequent stroke, heart failure, myocardial infarction, and mortality in newly-diagnosed AF patients in rural versus urban areas. Patients in rural areas were more likely to initiate anticoagulant treatment; however, they were less likely to initiate a newer class of anticoagulants compared to those in urban areas. Those in rural areas had modestly higher risk of cardiovascular outcomes and mortality compared to those in urban areas. Proteomics aids in understanding AF mechanisms and improves risk prediction. Future research should validate our prediction models, develop meaningful ways to incorporate protein biomarkers in clinical practice, and focus on improving AF treatment in rural areas.
  • Thumbnail Image
    Item
    Factors Influencing Beaver (Castor Canadensis) Population Fluctuations, And Their Ecological Relationship With Salmonids
    (2019-08) Johnson-Bice, Sean
    Within the western Great Lakes (WGL) U.S. region (Michigan, Minnesota, Wisconsin), the ecological impacts that North American beavers (Castor canadensis) have on cold- water streams are generally considered to negatively affect salmonid populations where the two taxa interact. Here, we review the history of beaver-salmonid interactions within the WGL region, describe how this relationship and management actions have evolved over the past century, and review all published studies from the region that have evaluated beaver-salmonid interactions. Our review suggests the impact beavers have varies spatially and temporally, depending on a variety of local ecological characteristics. We found beaver activity is often deleterious to salmonids in low-gradient stream basins, but generally beneficial in high- gradient basins; and ample groundwater inputs can offset the potential negative effects of beavers by stabilizing the hydrologic and thermal regimes within streams. However, there was an obvious lack of empirical data and/or experimental controls within the reviewed studies, which we suggest emphasizes the need for more data-driven beaver-salmonid research in the WGL region. Resource managers are routinely faced with an ecological dilemma between maintaining natural environmental processes within cold-water ecosystems and conducting beaver control for the benefit of salmonids, and this dilemma is further complicated when the salmonids in question are a non-native species. We anticipate future beaver-salmonid research will lead to a greater understanding of this ecologically-complex relationship that may better inform managers when and where beaver control is necessary to achieve the desired management objectives. Understanding how wildlife populations respond to density-dependent (DD) and density- independent (DI) factors is critically important for wildlife management and research, as this knowledge can allow us to predict population responses to forcing mechanisms such as climate, predation, and exploitation. Recent advancements in statistical methods have allowed researchers to disentangle the relative influence each factor has on wildlife population dynamics, but this work is ongoing. Using a long-term dataset collected from 1975 to 2002, we sought to evaluate the relative influence DD and a suite of covariates (weather, harvest, habitat quality, and wolf [Canis lupus] predation) had on annual rates of change in the number of beaver (Castor canadensis) colonies among 15 populations in northern Minnesota, USA. We modeled changes in beaver colony densities using a discrete-time Gompertz model within a Bayesian inference framework, and compared model performance among three global models using Deviance Information Criterion (DIC) widely available information criterion (WAIC): a DI model without covariates; a DD model without covariates; and a DD model with covariates. Our results provide strong evidence for compensatory (negative) DD within beaver colony dynamics. We found no evidence that covariates related to harvest, wolf predation, or habitat quality significantly influenced beaver colony growth rates, but cold winters (lag-0), spring drought (lag-0), and fall drought conditions (lag-2) were correlated with greater colony growth rates. Despite strong evidence of the effect of environmental covariates on beaver colony dynamics, prediction of colony dynamics using these covariates showed only minimal improvements. We suggest the lack of improvement in prediction was the result of model over-fitting, indicating our significant covariate effects may not be biologically relevant. Our analysis demonstrates how reliance on information criterion values may lead to erroneous conclusions in time-series analyses, and using a hindcasting approach like the one we present here may help determine whether model results are biologically relevant or merely statistically significant. Our results highlight the importance of long-term monitoring programs for evaluating the efficacy of predictive ecological models. That beaver populations are primarily intrinsically regulated has important management implications depending on whether the objectives concern eradicating beavers from unwanted regions, mitigating conflicts, or facilitating rewilding or colonization efforts.
  • Thumbnail Image
    Item
    Modeling Outputs of Efficient Compressibility Estimators
    (2018-06) Asamoah Owusu, Dennis
    There are times when it is helpful to know whether data is compressible before expending computational resources to compress it. The standard deviation of the byte distribution of data is an example of a measure of compressibility that does not involve actually compressing the data. This work considered five such measures of compressibility: byte standard deviation, shannon entropy, “average meaning entropy”, “byte counting” and “heuristic method”. We developed models that relate the output of these measures to the compression ratios of gzip, lz4 and xz using data retrieved from browsing Facebook, Wikipedia and YouTube. The models for byte standard deviation, shannon entropy and “average meaning entropy” were linear in both the parameters and the variables. The model for “byte counting” was non-linear in the predictor variable but linear in the parameters. The “heuristic method” was a classification model. In general, there was a strong relationship between the measures and the compressibility of a given data. Also, in many cases the models developed using one set of data from a source (like Youtube) was able to estimate the compressibility of another data set from the same source to a useful extent. This suggests the potential for developing a model per ECE for a source that can predict, to a useful degree, the compressibility of data from that source. At the same time, the differences in accuracy when models were evaluated on the data they were developed from versus when evaluated on new data from the same source indicate that there are important differences in the nature of the data coming from even the same source.
  • Thumbnail Image
    Item
    Numerical analysis of prediction with expert advice
    (2022-05) Mosaphir, Drisana
    This work investigates the online machine learning problem of prediction with expert advice through numerical analysis of a related PDE. The problem is a repeated two-person game involving decision-making at each step informed by n experts with geometric stopping condition; the continuum limit of the consequences of this game over a large number of steps leads to an elliptic PDE. This work presents a numerical scheme that allows us to solve this PDE for general number of experts n, and gives numerical results for n < 9.
  • Thumbnail Image
    Item
    R Code and Output Supporting: Used-habitat calibration plots: A new procedure for validating species distribution, resource selection, and step-selection models
    (2016-08-11) Fieberg, John R; Forester, James D; Street, Garrett M; Johnson, Douglas H; ArchMiller, Althea A; Matthiopoulos, Jason; jfieberg@umn.edu; Fieberg, John R
    Species distribution models (SDMs) are one of a variety of statistical methods that link individuals, populations, and species to the habitats they occupy. In Fieberg et al. "Used-habitat calibration plots: A new procedure for validating species distribution, resource selection, and step-selection models", we introduce a new method for model calibration, which we call Used-Habitat Calibration plots (UHC plots) that can be applied across the entire spectrum of SDMs. Here, we share the Program R code and data necessary to replicate all three of the examples from the manuscript that together demonstrate how UHC plots can help with three fundamental challenges of habitat modeling: identifying missing covariates, non-linearity, and multicollinearity.