Learning discrete decomposable graphical models via constraint optimization

Statistical model learning problems are traditionally solved using either heuristic greedy optimization or stochastic simulation, such as Markov chain Monte Carlo or simulated annealing. Recently, there has been an increasing interest in the use of combinatorial search methods, including those based on computational logic. Some of these methods are particularly attractive since they can also be successful in proving the global optimality of solutions, in contrast to stochastic algorithms that only guarantee optimality at the limit. Here we improve and generalize a recently introduced constraint-based method for learning undirected graphical models. The new method combines perfect elimination orderings with various strategies for solution pruning and offers a dramatic improvement both in terms of time and memory complexity. We also show that the method is capable of efficiently handling a more general class of models, called stratified/labeled graphical models, which have an astronomically larger model space.     

Die Anderen als sozialer Kontext. Zur Bedeutung strategischer Interaktion

An action is defined as “strategic” when the consequences of ego’s action depend on the action of alter. Situations of strategic interaction are numerous in daily life, business, and politics. Pioneers like Erving Goffman or Raymond Boudon recognized the importance of strategic interaction in sociological analysis long ago. Other peoples’ opportunities of actions form ego’s strategic context. The dynamics of the impact of the strategic context on ego’s action can be modeled and analyzed by means of game theory. We will discuss three examples of strategic interaction models: “Diffusion of responsibility”, Boudon’s “logic of relative frustration”, and the problem of social exchange and trust. We demonstrate the effects of the strategic context on the opportunities and beliefs of actors. In contrast to non-strategic rational choice theory, beliefs and opportunities are not assumed as exogenous. The analysis of the strategic context contributes to a better understanding of the micro-level effects and the macro-level implications. However, the strict rationality requirements of game models are often violated. In these situations, evolutionary models based on principles of learning and adaptions are more adequate than models based on assumptions of strict rationality.     

How Does the Individual Find a Place in French Sociology?

In France over the last four decades increasing interest in the small, the specific and the subjective can be observed. This involves a shift away from a holist approach towards social analysis focused on the individual and often carried out by researchers as specified and reflexive individuals. To be understood, this movement must be situated in the context of the development of French sociology, largely dominated at its origins by Durkheimian sociology. Thus, when the individual has appeared this is often not in his or her own right, but as representative of a social trend or group. As from the 1980s, new perspectives, very often imported from American sociology bring into play individual action, emotions and bodies, often studied as socially shaped. More recently, individual characteristics have been examined as a constraint on the possibilities of action, and discussion has also revolved around the reality of the person as an autonomous subject.     

The effect of accounting standards on Big Bath behavior

This paper investigates the influence of accounting standards on systematic earnings understatement. To isolate the effect of reporting regulations from those of other institutional drivers of earnings management, Holthausen (2003) proposes analyzing reporting practices in a country undergoing a transition in accounting standards. We respond to this call by analyzing earnings management practices of large German-listed companies before and after their transitions from German GAAP to IFRS. Specifically, we focus on Big Bath behavior following Chief Financial Officer turnovers between 1999 and 2006. Our findings provide evidence of the effect of accounting standards on the existence of this particular earnings understatement pattern. However, while we detect Big Bath behavior for our sample of German GAAP firms, we find no such indication for those firms reporting under IFRS. Controlling for alternative explanations such as self-selection bias does not change our findings. We conclude that accounting standards seem to have a considerable influence on earnings management behavior, independent of managers’ and auditors’ incentives.     

Education and Life Satisfaction in Relation to the Probability of Social Trust: a Conceptual Framework and Empirical Analysis

We employ cross-sectional microdata from 2010 through 2012 collected by the Italian National Institute of Statistics to investigate the proposed conceptual framework regarding the joint relationship between education, life satisfaction and the probability of social trust. The analysis has been carried out using two alternative specifications of the logit model: parametric and semiparametric. Our findings suggest that (a) both modelling methods that we utilise yield consistent results in terms of the positive effects of education and life satisfaction on social trust; (b) for unsatisfied individuals, social trust is generally low and largely unresponsive to the individual’s education level, whereas for ‘sufficiently satisfied’ individuals, social trust increases strongly with education; (c) there are some gender differences in the joint relationship between education, life satisfaction and social trust and (d) the parametric approach tends to mask some interesting patterns that are captured by the semiparametric specification. This finding implies that the parametric approach leads to a biased interpretation of the results, with an apparent failure of some of the assumptions made in the conceptual framework.     

Data-driven Kriging models based on FANOVA-decomposition

Kriging models have been widely used in computer experiments for the analysis of time-consuming computer codes. Based on kernels, they are flexible and can be tuned to many situations. In this paper, we construct kernels that reproduce the computer code complexity by mimicking its interaction structure. While the standard tensor-product kernel implicitly assumes that all interactions are active, the new kernels are suited for a general interaction structure, and will take advantage of the absence of interaction between some inputs. The methodology is twofold. First, the interaction structure is estimated from the data, using a first initial standard Kriging model, and represented by a so-called FANOVA graph. New FANOVA-based sensitivity indices are introduced to detect active interactions. Then this graph is used to derive the form of the kernel, and the corresponding Kriging model is estimated by maximum likelihood. The performance of the overall procedure is illustrated by several 3-dimensional and 6-dimensional simulated and real examples. A substantial improvement is observed when the computer code has a relatively high level of complexity.     

Determinants and effects of formal target agreements: an empirical investigation of German firms

We investigate organizational determinants and performance effects of formal target agreements for employees based on the IAB establishment panel, a representative panel of German establishments. The results show that establishments with a high employee turnover rate, a large proportion of temporary workers, and a highly qualified workforce are significantly more likely to implement target agreements. Also, establishments with works councils installed are more likely to do so, controlling, among others, for firm size. Concerning organizational performance, we apply a first-difference and a fixed effects approach and find that establishments that implement target agreements achieve around 5 % higher total sales compared to firms that did not introduce this practice. Hence, organizations seem to benefit from the implementation of formal target agreements for employees.     

Technische Softwaredokumentation unter ?konomischen Gesichtspunkten – Ein formal-deduktiver Ansatz

Only a very low priority seems to be given to software documentation. For example, about 80% of the software systems in production worldwide are badly documented. In contrast empirical studies confirm that an adequate technical documentation reduces the error rate of subsequent changes during software maintenance. This leads to the question about the optimal scale of investment in technical documentation in any software project. In research, this decision problem has been largely neglected. Since underlying cause-effect relationships may be empirically hard to detect, this paper proposes a new deductive approach, which compares the present value of payments for the documentation and bug fixing. The basis of this model is the phenomenon of error propagation from one component to adjacent components through its dependencies. As a result the following hypothesis can be inferred: It is always beneficial to document too much rather than comparatively too little even for an arbitrarily low quality of documentation and for any required rate of interest. This hypothesis shows robustness even against some changes in assumptions.