Πλοήγηση ανά Επιβλέπων "Kyriazidou, Ekaterini"

A B C D E F G H I J K L M N O P Q R S T U V W X Y Z

Α Β Γ Δ Ε Ζ Η Θ Ι Κ Λ Μ Ν Ξ Ο Π Ρ Σ Τ Υ Φ Χ Ψ Ω

Τώρα δείχνει 1 - 13 από 13

The agricultural productivity in Sub-Saharan Africa: an application of non-linear panel data models
Konstantinou, Adamantia; Athens University of Economics and Business, Department of Economics; Kyriazidou, Ekaterini
Despite the fact that sub-Saharan countries are characterised by high potentials in the agricultural sector, it seems that they do not exploit them. Even if many researchers recommend an increase in the use of chemical fertilizers so that higher agricultural productivity rates could be achieved, Africa is still behind in fertilizer use. In this dissertation, we try to explore the factors which determine the adoption and intensity of chemical fertilizer use. The analysis is based on a panel dataset from the Amhara region of Ethiopia. Both the farm characteristics and the socio-economic ones of the rural households are essential for taking the decision about the adoption and fertilizer use. The positive effects from the participation in MFIs and cooperatives and the negative effects from the credit constraints on the adoption, indicate the important role which the structure of rural credit markets and the access on credits play on the use of chemical fertilizers.
Credit scoring: retrospection, implementation today & application of fundamental econometric models on German Credit dataset
Ατζαμόγλου, Χαράλαμπος; Athens University of Economics and Business, Department of Economics; Palivos, Theodoros; Arvanitis, Stylianos; Kyriazidou, Ekaterini
One of the most important tools of the evaluation process of a customer’ s capability to pay off a loan and classify the customer into “bad risks” or “good risks” is credit scoring. An additional role of credit scoring is to reduce the possibility of a customer to default, which predicts the borrower’s risk level. The basic idea is to compare the characteristics of a customer with the characteristics of other customers of previous periods. If the customer ‘s characteristics are similar to those who have been granted credit and paid off the application will be approved. There are two problems in using credit scoring models for larger enterprises. The first one includes the information and its quality. As the size of the companies we examine is getting bigger, financial information is getting more important. Since companies are not obliged to keep records of their clients, the most of the characteristics of datasets refer to the owners’ repayment history than their financial status on their working activities The second obstacle also has to do with the information. The base of the problem is structural. Generally, analysts consider positive to test large populations whose members’ characteristics have a satisfying degree of homogeneity. As the market is consisted of a variety of sectors, it becomes difficult for analysts and agencies to collect homogeneous data of their clients. Taking into consideration everything mentioned above, this paper is going to present and analyze the main methods used in the credit scoring processes.
The determinants of NBA player salaries: a regression analysis
Margonis, Andreas; Athens University of Economics and Business, Department of Economics; Kyriazidou, Ekaterini
In this thesis an effort has been made to find the determinants of NBA players’ salaries. A unique dataset is consisted by panel data of contracts, personal characteristics and on-court performance statistics of 1895 NBA players, which are collected by hand, from 1998-99 to 2015-16 NBA season. Multiple regressions will be run including a shirking model so that the statistical significance and the sign of coefficients are found in order to observe the influence they have on the NBA players’ salaries as well as the time horizon, in which the factors that are considered by the NBA teams are consisted. In the last part, there is an attempt to examine if there is an exogenous time effect on the salaries of the NBA players with the use of the models which come of the research. The results show that NBA teams consider only the players’ performance from the season before they sign their contracts while there is no sign of salary discrimination based on nationality or position.
Dynamic panel data models: short & long T
Chatziathanasiou, Germanos Anargyros; Athens University of Economics and Business, Department of Economics; Kyriazidou, Ekaterini
Fixed effects estimator is a widely known estimator for panel data models whichcan be severely biased. The cause of this shortcoming is the so called incidentalparameter problem first stated by Neyman and Scott [14] for fixed T panels. Alvarezand Arellano [2] have shown that the asymptotic bias is reducing as T grows toinfinity and derived its asymptotic distribution. In practice growing T to infinityis not realistic assumption since most microeconomic panels have moderately smallT. Hence, a new approach was introduced by Hahn and Kuersteiner [6][5]whoconsidered alternative assumptions on the asymptotics, i.e that N, T grow at thesame rate, as an approximation which assist us on studying the bias term properties.As a result the bias reduction estimator is centered at the truth, whereas fixed effectsestimator is not. Due to the fact that bias reduction techniques require fixed effectsestimator to be already calculated, they are computationally heavy. While thismay be true, Monte Carlo studies have shown that they outperform many otherestimators while they keep the asymptotic efficiency.
Essays in applied microeconometrics : I. Discrete deforestation with spatial interactions II. Employment transitions of older persons in Britain III. Sibling post-schooling educational attainment
Emmanouilidi, Elpianna; Athens University of Economics and Business, Department of Economics; Kyriazidou, Ekaterini
Chapter one is an environmental application investigating land use changes of forests and semi-natural areas in the Greek region of Western Attica between 1990 and 2000. Its objective is to estimate the spatial equilibrium distribution of individual deforestation actions and determine the degree of coordination in individual behavior. For this purpose, a virtual economic network of 156 agents has been created by laying an ad hoc square grid over the region. Next, dominant forest land use changes have been determined in each land parcel using CORINE land cover maps for the years 1990 and 2000. The economic model is a discrete choice model with endogenous spatial interactions. Even though spatial interactions produce multiple equilibria, the present research proposes a two-stage fixed point estimator yielding a unique solution. Empirical findings suggest that equilibrium deforestation actions are strategic substitutes for the environment and complements for agriculture, and are characterized by a relative lack of coordination in individual behavior. Chapter two uses a dynamic panel in order to study the employment transitions of older persons in Britain in a discrete-time discrete-choice setting in order to identify both observable and unobservable factors which might influence both their labour market participation and status. For this purpose, a sample of 1097 individuals, drawn from the first five waves of the English Longitudinal Study of Ageing (ELSA), born 1946-1956 and aged 45-65 during the 10-year sampling period 2002-2011, is used. Three mutually exclusive employment states are considered with reference to non-participation, namely: full-time employment, part-time employment and self- employment. There is evidence that the 45-65 ELSA respondents under study exhibit significant mobility as they age. Their employment dynamics are characterized by significant positive first order state dependence in individual preferences for work. The long-run determinants of this unobserved heterogeneity indicate, among others, that self-perceived poor general health causes a negative selection out of the more flexible employment states while other household income has a negative impact for the more flexible employment states as opposed to full-time employment. The long- run effects are lower and asymmetric compared to the initial conditions. Next, the study investigates the determinants of the labour market participation of those sample respondents who appear in the sample along with their partner or spouse as well as the presence of any duration dependence in their choices. There is evidence that good as opposed to poor self-perceived health has a positive influence on participation irrespective of the partner's activity status while the effects are asymmetric and twice as high in case of poor health when the partner is inactive. Moreover, couple duration dependence is negative and increases with the length of the spell. Chapter three uses data regarding the children of the 1979 batch cohort of the National Longitudinal Survey of Youth (NLSY79) mothers, namely the Children and Young Adults (CYA) section, in order to investigate, on the one hand, the determinants of sibling post-schooling educational attainment, as measured by highest grade completion and college enrollment, and, on the other hand, the degree of resemblance therein. Empirical interest in sibling educational attainment and resemblance became systematic in the 1970s in the United States, when both sociologists and economists focused on the family as a collective economic decision unit, allocating resources in a competitive way which reduces inequality. In the present context, post-schooling educational achievement is identified by common family background factors as well as sibling own past and present characteristics. Attainment is examined using panel data with family-specific (mother) effects. Siblings are ordered in reverse birth order with a maximum birth lag of 10 years, hence the sibling index stands for the time dimension in a short panel. Next, first order reciprocal resemblance is examined using simultaneous equations methods for sibling data grouped on the mother, namely: 3SLS and endogenous Poisson estimation. Empirical evidence regarding post-schooling educational attainment indicates that sex and race have a positive effect apart from highest grade completion in case of Hispanics. In line with the empirical evidence, family income below the poverty level has a clear negative effect. Moreover, the childhood predictor of future academic achievement has a positive effect, as opposed to the behavioural problems scores. Similar to having experienced a young marriage, current behavioural risk, as measured by the heavy use of marijuana, has a clear negative effect. As far as first order reciprocal resemblance is concerned, estimates are positive but asymmetric and higher in the direction lag-one older to younger sibling, increasing with sibship size in case of linear system as opposed to endogenous Poisson estimation. In conclusion, there seems to be evidence in favour of resource dilution in poor families and confluence in non-poor families.
Essays in panel data and network econometrics
(13-09-2023) Χούντας, Κωνσταντίνος; Chountas, Konstantinos; Athens University of Economics and Business, Department of Economics; Arvanitis, Stylianos; Palivos, Theodoros; Dendramis, Yiannis; Christopoulos, Dimitrios; Ioannidis, Ioannis; Gkenakos, Xristos; Kyriazidou, Ekaterini
Μια αυξανόμενη βιβλιογραφία προσπαθεί να διακρίνει μεταξύ των παραγόντων που οδηγούν στην παρατηρούμενη συσχέτιση ανάμεσα στα αποτελέσματα των ατόμων, εστιάζοντας στον ρόλο των κοινωνικών δικτύων. Η παρούσα διατριβή επικεντρώνεται στις αλληλεπιδράσεις των ζευγαριών σχετικά με την απασχόληση και στις αλληλεπιδράσεις των αδελφών στην εκπαίδευση. Οι κύριες συνεισφορές είναι δύο: Πρώτον, προτείνεται ένα οικονομετρικό πλαίσιο μοντελοποίησης για την αναγνώριση των μεσολαβητικών διαύλων μέσω των οποίων αυτές οι αλληλεπιδράσεις λαμβάνουν χώρα· και δεύτερον, αναπτύσσονται κατάλληλα εργαλεία στατιστικής επαγωγής προκειμένου να εξεταστούν εμπειρικά οι ερωτήσεις που τίθενται. Το πρώτο κεφάλαιο στοχεύει στην εξήγηση της απόκλισης στα αποτελέσματα (όπως αποτελέσματα στην αγορά εργασίας ή στην εκπαίδευση) μεταξύ των αδελφών, συνδυάζοντας δύο πτυχές της βιβλιογραφίας: αυτή που χρησιμοποιεί τη σειρά γέννησης για να εξηγήσει αυτήν την απόκλιση και αυτή που μελετά τις αλληλεπιδράσεις μεταξύ των αδελφών. Η προτεινόμενη λύση είναι να υιοθετηθεί ένα ενιαίο οικονομετρικό πλαίσιο που να συμπεριλαμβάνει τόσο τις επιδράσεις της σειράς γέννησης όσο και τις επιδράσεις των αδελφών και να προχωρήσει στην ανάλυση του συνολικού αποτελέσματος της σειράς γέννησης, διακρίνοντας τον άμεσο αντίκτυπο από τον έμμεσο αντίκτυπο που οφείλεται στις αλληλεπιδράσεις των αδελφών. Το δεύτερο κεφάλαιο συμβάλλει στη βιβλιογραφία της οικονομετρίας των δικτύων και παρουσιάζει μία σειρά από αποτελέσματα που μπορούν να χρησιμοποιηθούν για την εκτίμηση του γραμμικού μοντέλου δικτύων, όταν δεν υπάρχουν διαθέσιμα δεδομένα που περιέχουν πληροφορίες για τις συνδέσεις μεταξύ των ατόμων που συμμετέχουν στο δίκτυο. Στο τρίτο κεφάλαιο, χρησιμοποιείται ένα διμεταβλητό μη-γραμμικό μοντέλο για τη μοντελοποίηση των μεταβάσεων των εργασιακών αποτελεσμάτων των ζευγαριών, λαμβάνοντας υπόψη τόσο την αλληλοσυσχέτιση στην εργασιακή κατάσταση του ζευγαριού όσο και την επίδραση των προγενέστερων εργασιακών καταστάσεων των μελών του ζευγαρίου.
Essays on applied panel data econometrics
(Athens University of Economics and Business, 12-2014) Polycarpou, Ioannis; Athens University of Economics and Business, Department of Economics; Genakos, Christos; Kyriazidou, Ekaterini
Doctoral thesis - Athens University of Economics and Business.
Model specification for football players' performance rating index
Koulitsi, Stavroula; Athens University of Economics and Business, Department of Economics; Kyriazidou, Ekaterini
The aim of this research is to specify and estimate models for football players' performance. Using these models, a rating index will be constructed for each player position, so that each players' contribution to the game can be comparable to one another. The position specialization is indispensable, since in each position the effectiveness of a player in special statistical categories is more important than in others. Each model will be a function of all the statistical measures that OLS regression will indicate as important. Although rating indices exist in other sports, such ratings do not exist in football. We will use the simulation video game Football Manager (FM), which provides the most updated real players database, as source of the needed data. In this game the user coaches a football team, dealing with transfers, training and game tactics, having as his main goal the teams’ success. Two seasons were played by the author and all the individual players’ statistics after each game, including the rating that FM provides, were collected by hand. The model selection, for each of the six positions, was performed via the combination of cross-validation technique, stepwise procedure and the use of information criteria. The models’ estimated coefficients were used, combined with the official players’ statistics of Tottenham Hotspur of the two previous seasons (English Premier League, FA Cup, League Cup games), in order to evaluate each player’s performance and identify the most valuable players. Furthermore, the impact of position performance to the outcome of the game was examined through ordered Probit and Logit analysis. Finally, the impact of the Tottenham Hotspur players’ season average ratings to their weekly wages and their market values was investigated.
On nonparametric and neural networks techniques for regression
Karoukis, Dimitrios; Athens University of Economics and Business, Department of Economics; Kyriazidou, Ekaterini
The context of this Thesis lies in the field of Econometrics. Our objective is to analyze various Nonparametric and Neural Networks techniques for regression. The first chapter of the Thesis is preoccupied with the analysis of the kernel density estimator, which is a fundamental step towards using kernel functions in regression. The second chapter is preoccupied with regression analysis by means of local polynomials. We will explore the techniques, their properties and their limitations. The third chapter is preoccupied with neural networks analysis. Specifically we present the structure of a one-hidden-layer and a two-hidden-layer feed forward neural network and explore their applications in regression. In the appendix we provide proofs for all the results that need to be validated throughout the thesis, namely, asymptotic (un)biasedness, consistency and asymptotic normality of the proposed estimators and the universal approximation theorem for neural networks in both the unit cube and the n-dimensional real space. We have used the R programming language for our analysis. We provide the algorithms in the appendix.
Prediction of S&P 500 index movement using data mining techniques
Michailidoy, Myrto-Christina; Athens University of Economics and Business, Department of Economics; Tzavalis, Elias; Arvanitis, Stylianos; Kyriazidou, Ekaterini
Predicting financial time series has proven extremely challenging due to their characteristics. There has been a vast number of researches investigating the predictability of financial variables from different aspects and by using different approaches. This study attempts to predict the direction of movement of S&P500 using models based on classification techniques; namely Logit, Linear Discriminant Analysis, Quadratic Discriminant Analysis, k-Nearest Neighbors, Support Vector Machines and Random Forest. The models developed are efficient, in the sense that any undetermined parameter is tuned using the Cross Validation technique. As inputs of the models, eleven technical indicators have been used and the data set is splitted into two sub-samples, the train and the test set. The performance of each model is assessed based on some evaluation measures, from which the best model is selected.
Prediction of stock market index movement using machine learning techniques
(28-02-2019) Impraimakis, Filippos; Athens University of Economics and Business, Department of Economics; Dimelis, Sophia; Sakellaris, Plutarchos; Kyriazidou, Ekaterini
It goes without saying that the ability to predict the direction of stock/index is of paramount importance for the viability of the companies and individual investors. An accurate prediction of the sign of a stock index is an effective hedging strategy that can mitigate the risk level of companies. In essence, setting the risk is a mean that can yield a more efficient allocation of the companies' capital. In this study, eight different classification techniques were employed for the determination of the direction of the S&P 500. Two different approaches were used as inputs, first for the acquired principal components generated from PCA and second for our existing dataset. This comparison showed that Principal Component Analysis (PCA) negatively affect our results, except the KNN algorithm. Our experimental results verified the superiority of Support Vector Machines (SVM) in predicting financial time series.
Statistical methods applied to credit scoring, using German credit data
(11/21/2018) Griva, Maria; Athens University of Economics and Business, Department of Accounting and Finance; Palivos, Theodoros; Dimelis, Sophia; Kyriazidou, Ekaterini
Statistical techniques consist credit decision-makers’ tools which are used by the banks and some companies to assess if the customers or loan applicants are capable to repay their obligations. In other words, if the customers are creditworthy or not (“good” or “bad”).In this thesis a German credit dataset is used and with the help of R programming we tried to examine six different statistical methods in order to make the best prediction. Different statistical-classification methods are performed and predict weather a loan applicant is creditworthy or not. Logistic Regression, Linear Discriminant Analysis, Quadratic Discriminant Analysis, K-Nearest Neighbors, Tree Based Methods: Classification trees (Decision Trees) and Random Forest are some statistical methods which are examined in this thesis.
The economics of happiness: a machine learning approach
(03/28/2018) Gavrill, Nikolaos; Γαβριήλ, Νικόλαος; Athens University of Economics and Business, Department of Economics; Vettas, Nikolaos; Palivos, Theodoros; Kyriazidou, Ekaterini
The emotional state of happiness has been for a long time studied mainly by thefield of psychology. Economists lately acknowledge the fact that self-reportedmeasurements of happiness should play a bigger role in policy and economictheory. In this study we use economic and development indicators to model therelationship between happiness and a selected subset of these indicators. Toproduce models of high accuracy and interpretability advanced methods from thefields of statistical learning and data mining are used. GDP per capita, GDP growthand unemployment are generally considered as indicators of high importance andthis study confirms that statement. Moreover we provide evidence on therelationship between happiness and variables that indicate life expectancy,immigration and ethical values like gender equality, although more researchshould be conducted in order to validate the relationship of the these indicatorsand the aforementioned fundamental variables.