Full-text resources of CEJSH and other databases are now available in the new Library of Science.
Visit https://bibliotekanauki.pl

Results found: 3

first rewind previous Page / 1 next fast forward last

Search results

Search:
in the keywords:  Regresja nieparametryczna
help Sort By:

help Limit search:
first rewind previous Page / 1 next fast forward last
EN
The advantage of the parametric regression models is the possibility of interpretation of the parameters of the regression model, i.e. to determine the direction and strength of the influence of predictors on the dependent variable. Unfortunately, in practice - the nonlinearity of the real processes, the influence of the phenomena with various probability distributions and a small number of observations limits the building of parametric models while the interpretation of non-parametric models is either impossible or very limited. Frequently such interpretation is useful in the specified range of variation. This may be a typical range of variation - for example, between the second and third quartiles, or a specific range due to the nature of the modeled phenomenon or process. It is difficult however, to build parametric models based only on the range of explanatory variables, because in this way we exclude observations giving additional knowledge into the model. The essence of this study is to enable the interpretation of non-parametric models through the creation of additional observations with these models in an interesting range of explanatory variables. These observations create secondary dataset used for the construction of a parametric model, which can now be interpreted. Presented investigations compare - using simulation - parametric models created for secondary sample with parametric models calculated for the original data.
PL
Artykuł jest poświęcony zagadnieniu odporności metod regresji na obserwacje odstające występujące w zbiorze danych. W pierwszej części przedstawiono wybrane metody identyfikacji obserwacji nietypowych. Następnie badano odporność trzech nieparametrycznych metod regresji: PPR, POLYMARS i RANDOM FORESTS. Analiz dokonano za pomocą procedur symulacyjnych na rzeczywistym zbiorze danych Mieszkania, w którym wykryto obserwacje odstające. Pomimo dosyć powszechnych przekonań o odporności regresji nieparametrycznej, okazało się, że modele zbudowane na całym zbiorze danych mają istotnie mniejsze zdolności predykcyjne niż modele uzyskane na zbiorze, z którego usunięto obserwacje nietypowe.
EN
The paper presents an important problem of robustness for outliers in regression. In the first part selected outliers detection techniques are described. Moreover, we empirically examine the robustness of the following methods: PPR, POLYMARS and RANDOM FORESTS on real world dataset. We show, that after removing outliers the prediction abilities of the models increase.
EN
In the paper the procedure for selecting the best nonparametric model for a given problem of regression is presented. This procedure has two stages. In the first one, many nonparametric models of regression, for different parameters settings, are built. Then the model with the smallest mean squared error is chosen. In the second stage, the method for the reduction of insignificant predictors is used. This procedure is applied to modelling household savings.
first rewind previous Page / 1 next fast forward last
JavaScript is turned off in your web browser. Turn it on to take full advantage of this site, then refresh the page.