[R] how to use normalmixEM to get correct result?

[Tatiana Benaglia]

Hi,
you're having problems with the initial values for the EM. If you  
don't give initial values for lambda, mu, and sigma, the function  
tries to generate them, which is causing the errors. I'd try to input  
initial values based on your data. Do the histogram of log.x (you  
don't need to sort the data) and choose values for mu and sigma. Then  
use:
outx = normalmixEM(log.x, lambda=c(.4, .6), mu=c(mu1, mu2),  sigma=c 
(sigma1, sigma2))

Hope that helps,
Tatiana


On Oct 14, 2007, at 11:00 PM, bgchen wrote:

> Dear R-Users,
>     I have a large number of data(54000) and the field of data is   
> 50 to 2.0e9. I want to use normalmixEM (package:mixtools) to fit  
> them in finite mixture narmal distributions,but get some mistakes.I  
> don't know which steps make the error.
> 	I have used the following functions before
> 	>x<-read.table("data")
> 	>log.x<-log10(x$V1)
>> log.x<-sort(log.x)
>> outx = normalmixEM(log.x,arbmean = TRUE, arbvar = TRUE,  
>> k=2,epsilon = 1e-03,maxit = 10000, verb = FALSE)
> 	The output is given as:
> 	Need new starting values due to singularity...
> 	Need new starting values due to singularity...
>     Need new starting values due to singularity...
>     error in normalmixEM(log.x, arbmean = TRUE, arbvar = TRUE, k =  
> 2, epsilon = 0.001,  :
>         Too many tries!
>
>
> Thanks for the help
>
>  				
>
> 　　　　　　　  bob
> 　　　　　　　　bgchen at scut.edu.cn
> 　　　　　　　　　　2007-10-15
> ______________________________________________
> R-help at r-project.org mailing list
> https://stat.ethz.ch/mailman/listinfo/r-help
> PLEASE do read the posting guide http://www.R-project.org/posting- 
> guide.html
> and provide commented, minimal, self-contained, reproducible code.