Browsing by Subject "Regression"

Now showing 1 - 6 of 6
  • Thumbnail Image
    ItemOpen Access
    A multilevel model of field-scale nitrogen export from agricultural areas
    (2010-04-27T00:00:30Z) Jones, Phillip
    Agricultural lands contribute significant nitrogen loads to surface waters. Excessive nitrogen input leads to eutrophication, the process by which aquatic ecosystems become nutrient rich. Eutrophication is associated with a wide range of undesirable changes, including shifts in physical and chemical states, changes in species composition, and the loss of ecosystem services. In agricultural areas, excessive nutrient loading is addressed through the implementation of Best Management Practices (BMPs). However, field-scale nutrient export is controlled by a complex array of interacting factors that operate at different spatial scales. Multilevel regression is a statistical technique that allows for the exploration of group-level factors that may explain variation in the overall model coefficients. In this study, multilevel regression models for dissolved and particulate nitrogen loading are fit to USDA agricultural data. The results indicate that the impact of management practice depends on the form of nitrogen as well as predictors such as soil texture that operate on large spatial scales. Specific management recommendations include soil nitrogen testing and the use of conservation measures that address water runoff. Management applications of the fitted models include load estimation as part of watershed leveling modeling efforts as well as the evaluation of proposed policy guidelines for nutrient control.
  • Thumbnail Image
    ItemOpen Access
    Civil Resistance or Rebellion: The Impact of Country-Level Factors on Revolutionary Strategy
    (2013) Carrington, Christopher

    This paper constitutes a partial answer to the question of when political

    resistance campaigns that use primarily violent or nonviolent strategies occur. In doing so, it attempts to bridge the gap between discussions of rebellion and civil resistance. A number of broad theoretical propositions are made and statistically tested by combining the NAVCO data on violent and nonviolent resistance campaigns with data that is commonly used in the civil war literature. The study finds that revolutionary civil resistance campaigns are unlikely to occur in democracies, population size does not obstruct nonviolent collective action, and the present favors nonviolent resistance more than the past, likely due to technological factors. It also provides evidence that divided societies are associated with rebellion rather than civil resistance.

  • Thumbnail Image
    ItemOpen Access
    Military Service and Civilian Labor Market Outcomes: Comparing Employment of Post-9/11 Veterans and Nonveterans
    (2016-01-05) Ordway, Matthew
    Veterans struggle to enter the civilian labor market following military service. Since the September 11 terrorist attacks, over 3.2 million Americans have served in the military. Upon returning home, these veterans are twenty percent more likely to be unemployed than nonveterans (7.2% vs 6%, respectively). This study investigates the association between military service and employment outcomes (employment status and weekly earnings) for post-9/11 veterans, a heretofore understudied group. Data was obtained from the Current Population Survey (CPS) Veteran Supplement. Linear probability models and OLS regressions were utilized to compare employment outcomes between veterans and nonveterans of similar age, education and race/ethnicity (“veteran effect”). Findings suggest that the veteran effect on employment is negative while the veteran effect on earnings, given employment, is positive. This is likely because of selection bias; the most productive veterans find employment and therefore command higher wages. Veteran effects differ by race and ethnicity, length of military service and time since service. Policymakers should tailor transition programs to the most vulnerable veterans, such as long-term military personnel.
  • Thumbnail Image
    ItemEmbargo
    Predicting All-cause Mortality among Chinese Community-Dwelling Elderly
    (2020) Jin, Xurui

    Background and aim: This study aimed at building the prediction model of all-cause mortality among Chinese dwelling elderly with different methods including regression models and machine learning models and to compare the performance of machine learning models with regression models on predicting mortality. Additionally, this study also aimed at ranking the predictors of mortality within different models and comparing the predictive value of different groups of predictors using the model with the best performance.Method: I used data from the Healthy Ageing and Biomarkers Cohort Study, a sub-study of the Chinese Longitudinal Healthy Longevity Survey (CLHLS). The baseline survey was conducted in 2008 and participants were followed every 2-3 years till 2018. The analysis sample included 2,448 participants. I used totally 117 predictors to build the prediction model, including 65 questionnaires, 39 biomarkers, and 15 genetics predictors. Four models were built (XG-Boost, random survival forest [RSF], Cox regression with all variables, and Cox-backward). I used C-index and integrated Brier score to evaluate the performance of those four models. Results: The XG-Boost model and RSF model shows slightly better predictive performance than Cox models and Cox-backward models based on the C-index and integrated Brier score. Age. The activity of daily living and Mini-Mental State Examination score were identified as the top 3 predictors in the XG-Boost and RSF models. Biomarker and questionnaire predictors have a similar predictive value, while genetic predictors have no addictive predictive value when combined with questionnaire or biomarker predictors. Conclusion: In this work, it is shown that machine learning techniques can be a useful tool for both prediction and its performance sightly outperformed the regression model in predicting survival.

  • Thumbnail Image
    ItemOpen Access
    Stumpage Price Impacts on Carbon Accumulation
    (2021-04-30) Ruan, Leyi; Scofield, Evan
    California’s Cap-and-Trade program, created by the California Air Resources Board (CARB), allows large-scale polluters in the state to partially offset their greenhouse gas emissions by purchasing carbon offset credits. Although there are several types of allowable credits, by far the most common are credits generated by forest carbon projects. These projects require landowners to demonstrate that their forests are storing carbon at levels that exceed what CARB has determined to be the Common Practice Baseline for the associated ecological region and forest type. CARB does not disclose its procedure for calculating these baselines, but it is known to use data collected by the U.S. Forest Service’s Forest Inventory and Analysis (FIA) Program. Forest carbon offset developers like our client, Bluesource LLC, rely on the CARB baselines and data on forest carbon stocking levels to identify potential development opportunities. In this report, we seek to determine if CARB baselines and FIA carbon levels can be predicted by prices received by forest owners for their timber, otherwise known as stumpage prices. Previous studies have demonstrated a theoretical effect of stumpage prices on timber harvests, hence on aboveground forest carbon levels, but none has modeled the empirical relationship between stumpage prices and either the CARB baselines or FIA carbon levels. The first section of this report provides an overview of the carbon market in California, focusing on how forest carbon offsets are developed by third-party organizations like Bluesource. This section also states our three research questions: 1. Can stumpage prices be used to explain CARB baselines? 2. Can stumpage prices be used to explain FIA aboveground carbon levels? 3. If the answers to questions 1 and 2 are affirmative, can the associated statistical models be used to predict future CARB baselines and FIA carbon levels? The second section describes the work done to collect data on stumpage prices from Maine, Michigan, and West Virginia, the CARB baseline report from 2011 and 2015, and the FIA carbon data from the same three states. There were significant differences in the reporting styles of the various datasets, so this section also explains how the data were cleaned and reformatted in preparation for our analyses. It provides as well some descriptive statistics to orient the reader on data characteristics. The third section describes the methods undertaken in our analyses. We used regression models to examine the relationship between stumpage prices and both the CARB baselines and the FIA carbon levels. Because the CARB and FIA data differ in the years in which they are reported, so to do the analysis methods. The CARB analysis utilized a cross-sectional regression, whereas the FIA analysis utilized a panel analysis. Beyond stumpage prices, we tested the effects of a number of other explanatory variables in each analysis and ultimately selected two models to be used to project future CARB baselines and FIA carbon levels. The model results and the results of their projections are also included in this section and are shown using maps of the expected percentage increase for each geographic region of our study. The fourth and final section highlights the implications of our findings both to our client and to the relevant literature. We also acknowledge limitations of our models and offer suggestions of how the models could be improved and adjusted for use in other geographic regions. Although stumpage prices influence forest owner harvesting behavior and thus forest carbon levels and CARB baselines too, they likely do not capture all forces influencing landowners’ forest management decisions. A more sophisticated model with more complete data may have greater predictive capability. This report makes several key points and recommendations: • Stumpage prices can be used as predictors of both CARB baselines and FIA carbon levels. • The models we developed could be improved by including more complete stumpage price data from the study region, including sales from both public and private forests instead of only one or the other. • The effects of other variables should be investigated, particularly interest rates and price trends (vs. price levels). • Analyzing the effect of FIA carbon levels on CARB baselines may provide a more direct way to project future CARB baselines. • These models are applicable only to Maine, Michigan, and West Virginia, but the methods used can be replicated for any region with similar data.
  • Thumbnail Image
    ItemRestricted
    The Use of Error Components Models in Combining Cross Section with Time Series Data
    (1969) Wallace, Dudley; Hussain, Ashiq
    A mixed model of regression with error components is proposed as one of possible interest for combining cross section and time series data. For known variances, it is shown that Aitken estimators and covariance estimators are in one sense asymptotically equivalent, even though the Aitken estimators are more efficient in small samples. Turning to unknown variance components, Zellner-type iterative estimators are compared with covariance estimators. Here, few small sample properties are obtained. However, it is shown that covariance and Zellner-type estimators have equivalent asymptotic distributions and equivalent limits of sequences of first and second order moments for weakly nonstochastic regressors. For the model analyzed, the theoretical results obtained, as well as ease of computation, tend to support traditional covariance estimators of the regression parameters. An additional interesting result presented in an appendix is that ordinary least squares estimates of the β's (ignoring the error components) have unbounded asymptotic variances. On efficiency grounds, this argues rather strongly for some care in combining data from alternative sources in regression analysis.