|
Abstract:
|
In this talk, we present details on the mixturegram, a recently-developed visualization tool for assessing the number of components in finite mixture models. Our work provides a quick assessment for scientific researchers and supplements traditional numerical approaches for the problem of determining the number of components in a mixture model. The construction of the mixtruegram involves dimension reduction, and different techniques can be employed to improve the performance of the mixturegram. Therefore, our tool is adaptive to the development of new dimension reduction methods. Although the interpretation of the mixturegram is subjective, which is common for information graphics, we provide an ad hoc selection criterion based on calculations used for the mixturegram. We demonstrate the efficacy of this new tool using two real data applications, a univariate astronomy dataset and a multivariate cancer dataset. Our conclusions are further supported by the results obtained using information criteria.
|
