Suchergebnisse

AbstractThis article evaluates the reliability of sensitivity tests. Using Monte Carlo methods we show that, first, the definition of robustness exerts a large influence on the robustness of variables. Second and more importantly, our results also demonstrate that inferences based on sensitivity tests are most likely to be valid if determinants and confounders are almost uncorrelated and if the variables included in the true model exert a strong influence on outcomes. Third, no definition of robustness reliably avoids both false positives and false negatives. We find that for a wide variety of data-generating processes, rarely used definitions of robustness perform better than the frequently used model averaging rule suggested by Sala-i-Martin. Fourth, our results also suggest that Leamer's extreme bounds analysis and Bayesian model averaging are extremely unlikely to generate false positives. Thus, if based on these inferential criteria a variable is robust, it is almost certain to belong into the empirical model. Fifth and finally, we also show that researchers should avoid drawing inferences based on lack of robustness.

ABSTRACTThis paper explores linear programming‐like sensitivity analysis in decision theory. In particular, the paper considers the sensitivity of an optimal decision to changes in probabilities of the states of nature and the development of "confidence spheres" to bound arbitrary parametric changes in the probability vector. Such information can be used to assess the accuracy required in assigning probabilities and the confidence in the maximumutility decision.

We review briefly some examples that would support an extended role for quantitative sensitivity analysis in the context of model‐based analysis (Section 1). We then review what features a quantitative sensitivity analysis needs to have to play such a role (Section 2). The methods that meet these requirements are described in Section 3; an example is provided in Section 4. Some pointers to further research are set out in Section 5.

Discusses the methods of sensitivity analysis in use generally and
by the property appraisal profession. Proposes a simplified structured
and systematic technique of selecting critical or sensitive factors for
sensitivity analysis in property development and investment appraisal.
Concludes that sensitivity analysis has become an integral part of
property appraisal.

Decision analysts use sensitivity analysis to identify influential variables, to determine which input variables to model stochastically, and to characterize scenarios that could affect a change in the rank ordering of the alternatives. A frequently recommended sensitivity analysis technique is "one‐way" sensitivity analysis, which determines a variable's influence by the degree to which the objective function changes as that variable is varied while all other variables are held fixed. Disadvantages of one‐way analysis are that it measures the influence of only one variable at a time and it assumes independence among the input variables. Clearly, however, there are situations when dependencies exist among the input variables that could possibly affect the sensitivity analysis results. This research develops a strategy that incorporates dependence relations among the input variables into the sensitivity analysis using rank correlations. Only decision problems with a finite number of alternatives and continuous state variables are considered.

ABSTRACTThe tolerance approach to sensitivity analysis in linear programming provides a framework for dealing with simultaneous and independent variations of the coefficients and terms. Specifically, the approach yields a maximum tolerance percentage which in turn characterizes a region of variability for the coefficients or terms. This paper shows how to expand this region of variability yielding ordinary sensitivity analysis as a special case of the tolerance approach.

AbstractSurvey weighting allows researchers to account for bias in survey samples, due to unit nonresponse or convenience sampling, using measured demographic covariates. Unfortunately, in practice, it is impossible to know whether the estimated survey weights are sufficient to alleviate concerns about bias due to unobserved confounders or incorrect functional forms used in weighting. In the following paper, we propose two sensitivity analyses for the exclusion of important covariates: (1) a sensitivity analysis for partially observed confounders (i.e., variables measured across the survey sample, but not the target population) and (2) a sensitivity analysis for fully unobserved confounders (i.e., variables not measured in either the survey or the target population). We provide graphical and numerical summaries of the potential bias that arises from such confounders, and introduce a benchmarking approach that allows researchers to quantitatively reason about the sensitivity of their results. We demonstrate our proposed sensitivity analyses using state-level 2020 U.S. Presidential Election polls.

In a quantitative model with uncertain inputs, the uncertainty of the output can be summarized by a risk measure. We propose a sensitivity analysis method based on derivatives of the output risk measure, in the direction of model inputs. This produces a global sensitivity measure, explicitly linking sensitivity and uncertainty analyses. We focus on the case of distortion risk measures, defined as weighted averages of output percentiles, and prove a representation of the sensitivity measure that can be evaluated on a Monte Carlo sample, as a weighted average of gradients over the input space. When the analytical model is unknown or hard to work with, nonparametric techniques are used for gradient estimation. This process is demonstrated through the example of a nonlinear insurance loss model. Furthermore, the proposed framework is extended in order to measure sensitivity to constant model parameters, uncertain statistical parameters, and random factors driving dependence between model inputs.

Agent-based computational economics (ACE) is gaining interest in macroeconomic research. Agent-based models (ABM) are increasingly able to replicate micro- and macroeconomic stylised facts and to extend the knowledge about real-world economic systems. These advances allow ABM to become a valuable and more frequently used tool for policy analysis in academia and economic practice. However, ACE is a rather complex approach to already complex investigations like policy analyses, i.e. the analyses on how a variety of policy measures affects the (model) economy, which makes policy analyses in ABM prone to critique. The following research paper addresses these problems. We have developed a procedure for policy experiments in ACE which helps to conceptualise and conduct policy experiments in macroeconomic ABM efficiently. The procedure makes policy implementation decisions and their consequences transparent by conducting what we term the policy implementation sensitivity analysis (POSA). The application of the procedure produces graphical and/or numerical reports that should be included in the appendix of the original research paper in order to increase the credibility of the research, similar to proofs and protocols in analytical and empirical research. ; Agentenbasierte Modellierung (ABM) gewinnt in der makroökonomischen Forschung zunehmend an Bedeutung. Diese Modelle sind immer mehr in der Lage, mikro- und makroökonomische Muster aus der realen Welt abzubilden und das Wissen über wirtschaftliche Zusammenhänge zu erweitern. Dank dieser Fortschritte werden agentenbasierte Modelle zu einem wertvollen und immer häufiger genutzten Instrument der Policy-Analyse. Policy-Analyse meint in diesem Zusammenhang die Untersuchung, wie sich Politikmaßnahmen auf relevante (wirtschaftliche) Kennziffern (z.B. Arbeitslosigkeit oder Wirtschaftswachstum) auswirken. Allerdings ist agentenbasierte Modellierung eine komplexe Untersuchungsmethode für die ohnehin komplexe Policy-Analyse, da eine Politikmaßnahme zumeist ein Bündel aus verschiedenen Einzelmaßnahmen darstellt. Das macht Policy-Analysen, die auf Simulationsstudien mit agentenbasierten Modellen basieren, anfällig für Kritik. Es wird beispielsweise kritisiert, dass der Effekt einer Politikmaßnahme im Model bedingt ist durch die Wahl der Parameter, die diese Politikmaßnahme beschreiben (d.h. der Intensität der Maßnahme). Der folgende Beitrag greift diese Probleme auf. Wir haben ein Verfahren entwickelt, das hilft, die Simulationen von Politikmaßnahmen in verschiedenen Intensitätsgraden in makroökonomischen ABM zu konzeptualisieren und effizient durchzuführen. Wir nennen es 'Policy Implementation Sensitivity Analysis' (POSA). Das Verfahren macht sowohl die Beweggründe für die Implementierung einer Politikmaßnahme in einer bestimmten Intensität als auch die Konsequenzen dieser Entscheidung transparent. Die Anwendung von POSA führt zu grafischen Darstellungen und/oder statistischen Berichten, die im Anhang der eigentlichen Forschungsarbeit aufgenommen oder als Begleitmaterial online zur Verfügung gestellt werden können. Diese können die Glaubwürdigkeit der Forschung erhöhen, vergleichbar mit Protokollen in der analytischen und empirischen Wirtschaftsforschung. Im Anhang dieses Beitrags wird die Anwendung des POSA-Verfahrens anhand eines Beispiels veranschaulicht.

AbstractThe robustness of the assigned prior distribution in a Bayesian estimation problem is examined. A Bayesian analysis for a stochastic intensity parameter of a Poisson distribution is summarized in which the natural conjugate is assigned as the prior distribution of the random parameter. The sensitivity analysis is carried out by assuming the existence of a true prior which is different in form from that of the assigned prior distribution. By using mean‐squared error as a measure of performance, the ensuing Bayes decision function is compared to the corresponding minimum variance unbiased estimator. Results indicate that the Bayes estimator is largely robust to deviations from the assigned prior and remains squared‐error superior to the MVU type within a broad region.

AbstractDespite high degrees of uncertainty associated with graphical functions, sensitivity analysis of these functions has received less attention than parametric sensitivity analysis. Recently, a promising method was proposed in the literature to generate alternative functional forms, reducing the problem to that of parametric sensitivity. Yet the usability of the method for graphical functions in system dynamics has not been investigated. We apply this function distortion method to a sample model and identify a number of shortcomings, such as a limited variety of alternative forms. We then propose extensions to the method to address the shortcomings, and subsequently test the extensions. We find the (extended) method of function distortion to be readily applicable and efficient in testing the sensitivity of model outputs to variations in the form of graphical functions. The proposed extensions increase the variety of possible distortions, but further research can be conducted on the control of the distortions. Copyright © 2014 System Dynamics Society