1: \begin{abstract}

2: Estimation of the number of components (or order) of 

3: a finite mixture model is a long standing and challenging problem in statistics.

4: We propose the Group-Sort-Fuse (GSF) procedure---a new penalized likelihood approach

5: for simultaneous estimation of 

6: the order and mixing

7: measure in multidimensional finite mixture models. 

8: Unlike methods which fit and compare mixtures with varying 

9: orders 

10: using criteria involving model complexity, our approach directly penalizes a 

11: continuous function of the model parameters. More specifically,

12: given a conservative upper bound on the order, 

13: the \name groups and sorts mixture component parameters 

14: to fuse those which are redundant. 

15: For a wide range of finite mixture models, we show that

16: the \name is consistent in estimating the true 

17: mixture order

18: and achieves the $n^{-1/2}$ convergence 

19: rate for parameter estimation up to polylogarithmic factors. 

20: The GSF is implemented for several univariate and multivariate

21: mixture models in the R package \package. 

22: Its finite sample performance is supported by a

23: thorough simulation study, and its application

24: is illustrated on two real data examples.

25: \end{abstract}