When we fit a distribution of a sample in a classification problem with multiple groups, we will get more than one peak in the distribution plot. So, it would be difficult to identify which distribution function really fit the sample.
The reason to fit a distribution to a set of data is to parameterize the data in order to simplify calculation. The parameterized distribution is not the data and differences from the actual data distribution could be important. The parameters of the distribution can have important physical interpretations. A bimodal distribution often is an indication that separate processes produced the distribution. The parameterization of the data requires that the resulting distribution has a physical basis, two processes, separately parameterized, combine in a single, but resolvable bimodal distribution.
Multiple groups should have a distribution for each group. Each group should have parameters characteristic to each group. A bimodal or multimodal distribution indicates that the parameters can be distinguished.
The type of data should indicate the expected distribution for a group. Parameters should be selected for each group. The distribution function is the combined group distributions.
The reason to fit a distribution to a set of data is to parameterize the data in order to simplify calculation. The parameterized distribution is not the data and differences from the actual data distribution could be important. The parameters of the distribution can have important physical interpretations. A bimodal distribution often is an indication that separate processes produced the distribution. The parameterization of the data requires that the resulting distribution has a physical basis, two processes, separately parameterized, combine in a single, but resolvable bimodal distribution.
Multiple groups should have a distribution for each group. Each group should have parameters characteristic to each group. A bimodal or multimodal distribution indicates that the parameters can be distinguished.
The type of data should indicate the expected distribution for a group. Parameters should be selected for each group. The distribution function is the combined group distributions.
multimodal distribution can be tried. we have a worked out multimodal extensions of Logistic ,Laplace and Normal distribution.
Thank you very much Joseph Alvarez and Subrata Chakraborty for your answers. I need to learn more about what have been forwarded by Joseph Alvarez. Thanks again.
- Badges
- Science topic