EURO-Online login
- New to EURO? Create an account
- I forgot my username and/or my password.
- Help with cookies
(important for IE8 users)
1749. Probability models based on a mixture of densities
Invited abstract in session TB-31: Analytics and the link with stochastic dynamics II, stream Analytics.
Tuesday, 10:30-12:00Room: 046 (building: 208)
Authors (first author is the speaker)
| 1. | Lubos Marek
|
| Statistics and Probability, University of Economics, Prague |
Abstract
When we build the model of probability distributions, we can work with one distribution or use a mixture of distributions. In general practice, one density distribution is usually used. We will show another approach, using a mixture of distributions. The whole process will be demonstrated on wage data from the Czech Republic. For the construction of mixtures, it is possible to use densities of classical distributions for modelling wage data such as normal, logarithmic normal, Johnson´s and other distributions. When creating the final mixture, we solve two problems. First, we need to determine the number of components in the mixture. Secondly, we need to estimate the parameters of individual densities in the mixture. Since we have a large amount of data broken down by different criteria (age, gender, region, education, etc.), we can naturally divide the data into a large number of categories. Each category is then represented by one component in the resulting mixture. Detailed data breakdown allows us to calculate weights of individual components in the resulting mixture. In addition to the classic procedure known from theory, the authors in the article also offer a non-standard approach, which allows not only to model the course of past data, but also the possibility of estimating the future development of the wage distribution. In the conclusion, attention is also paid to the stability of the models.
Keywords
- Economic Modeling
- Forecasting
- Optimization Modeling
Status: accepted
Back to the list of papers