Vol: 32 Page: 241

Authors:
Ewa Bakinowska
Radosław Kala

Title:
 Estimation methods in generalized linear models

Language: Polish

Keywords:
generalized linear model
maximum likelihood method
least squares method
weighted least squares method

Summary:
Categorical response data are usually modeled with the use of the multinomial distribution. If the expected value of a random variable of this distribution is mapped by logistic transformation, we get the model, which belongs to the class of generalized linear model. Its unknown parameters can be obtained adopting various approaches. There are the geometrical methods such as the leas squares and the weighted least squares, and the iterative methods as the Fisher’s scoring and the Newton-Raphson. Apart from the model presentation and discussion of estimation methods, we pointed out the links among the estimation methods.