Οι οριακές επιδράσεις των επεξηγηματικών μεταβλητών ενός γενικευμένου γραμμικού μοντέλου: ερμηνεία και εφαρμογές

Γούτος, Κωνσταντίνος

Master Thesis

Author

Γούτος, Κωνσταντίνος

Date

2021-07

Abstract

Regression analysis is one of the most essential tools in modern Statistics. The estimates of the coefficients of explanatory variables are calculated to describe potential relationships in multivariate data and make predictions. For the classic Linear Regression Models, the interpretation of the estimates is straight forward. However, the regression framework is so wide that it allows to consider more complex relationships, including nonlinear relationships between the explanatory and the response variables. This provides the flexibility to identify potential complex relationship between many variables. On the other hand, there is a risk of misinterpretation. The purpose of this M.Sc. Thesis is the presentation, interpretation and application of the Marginal Effects, with a focus on the generalized linear models. Marginal Effects is an important tool for Statistical inference as they provide a measure of calculating the effect of a change of an explanatory variable on the response variable, where this change is measured in its physical units. This M.Sc. Thesis consists of 6 Chapters. In Chapter 1, we present the generalized linear models, emphasizing on the models for binary outcomes (logit, probit and complementary log-log) and for count outcomes (Poisson and zero inflated Poisson). In chapter 2, we introduce the Marginal effects and we describe the ways of calculating them. Chapter 3 consists of a detailed overview of the various approaches in which the Marginal Effects can be calculated with. Also, it includes the comparison for two of the presented methods. In Chapter 4, we describe the methods used to estimate the asymptotic standard errors of the Marginal Effects, emphasizing on the Delta method. In Chapter 5, we present the simplified models of Marginal Effects for the generalized linear models logit and probit after normalization of the explanatory variables. In Chapter 6, we apply the theoretical foundation of Marginal Effects. For the Case Study, a dataset related to a rare species of a crab, called horseshoe crab (Limulus polyphemus), is used ending up to remarkable conclusions. For the computational part of this thesis, we used the R Programming Language.

Postgraduate Studies Programme

Εφαρμοσμένη Στατιστική

Department

Σχολή Χρηματοοικονομικής και Στατιστικής. Τμήμα Στατιστικής και Ασφαλιστικής Επιστήμης

Number of pages

131

Language

Greek

URI

https://dione.lib.unipi.gr/xmlui/handle/unipi/13574
http://dx.doi.org/10.26267/unipi_dione/997

Collections

Τμήμα Στατιστικής και Ασφαλιστικής Επιστήμης

Show full item record

Except where otherwise noted, this item's license is described as
Αναφορά Δημιουργού-Μη Εμπορική Χρήση-Όχι Παράγωγα Έργα 3.0 Ελλάδα