• Hızlı Erişim
  • Announcements
  • DUYURU 3

    Yayın politikamız

    gereği,

    yayinlanacak

    makale sayısı

    sınırlandırılmış ve

    çalışmaların geliş

    tarihi itibarıyla

    yayınlanması

    kararlaştırılmıştır.


    DUYURU 2

    Dergimiz ana

    sayfasında

    "Dergi Hakkında"

    kısmında belirtilen

    alanlar dışında

    yayın kabul 

    etmeyecektir.

     


    DUYURU 1

    Çok yazarlı

    makalelerdeki

    tüm yazarların 

    dergiye 

    üye olması

    gerekmektedir.

    Aksi takdirde

    yazılar

    yayınlanmayacaktır.

     


Abstract


A ALTERNATIVE TO ORDINARY LEAST SQUARES REGRESSION: QUANTILE REGRESSION
Quantile Regression is a regression method which suggested by Koenker and Basett. Quantile regression is a form of robust (outlier resistant) regression. Basic regression models or Least Square methods are not flexible to outlier values. Basic regression methods or estimators are affected by extreme values and Quantile Regression is less sensitive to extreme values than the other regression model. Therefore, Quantile Regression uses in a range of application settings. Least Squares methods has some assumptions about variance of errors. Quantile Regression approach is advanced like an alternative to the Least Squares method used for regression analysis assumptions. Quantile regression is a way to estimate the conditional quantiles of a dependent variables distribution in the linear model. Quantile Regressions are effective for visualizing changes in the conditional distribution of data sets. Quantile Regression approach useful in such cases where especially the outlier values. The purpose of this study is to retest the effectiveness of the Quantile Regression method. For this purpose, synthetic data generated with the R program. There are dependent and independent parameters in dataset which contains extreme values. After synthetic dataset which exhibits normal distribution data has been generated, number of data was limited to 400. The reason why synthetic data is generated that it is thought that generating synthetic data will explain the theory better in theoretical studies. In the data exhibiting normal distribution, after adding the extreme values, Least Squares method and Quantile Regression method were compared. In the end of the simulation practice, according to RMSE criteria value, it is found that the performance of the Quantile Regression model is better than Least Squares model in the case of extreme values.

Keywords
Quantile Regression, least squares method, outlier values


Adres :ANKARA
Telefon :0506 509 32 59 Faks :
Eposta :info@avrasyad.com

Web Yazılım & Programlama Han Yazılım Bilişim Hizmetleri