R help - Feb 2008 - newbie (me) needs to model distribution as two overlapping gaussians

Recently, I have been working with some data that look like two overlapping
gaussian distributions.  I would like to either

1) determine the mean and SD for each of the two distributions

OR

2) get some (bayesian ?) statistic that estimates how likely an observation is
to belong to the left-hand or right-hand distribution

In case I'm using the wrong language, my data looks something like this:

B <- rnorm(500,40,10)
H <- rnorm(500,80,5 )
N <- runif(200,0,99)
D <- c(B,H,N)

Where B=background, H=hits, N=noise, and D=my observed distribution

I have seen analyses like this in the past, but I can't remember what it is
called.  If somebody out there can point me towards an R function, or even the
cannonical name for this kind of model, I think I can write the necessary code.

Thanks in advance,
Mark

Is a finite mixture of 2 gaussians the name you are looking for?
This specific model will not deal with your N component however.

you can find some functions here:
http://cran.r-project.org/web/views/Cluster.html

Thomas

mmiller at mail.usp.edu wrote:
> Recently, I have been working with some data that look like two overlapping
gaussian distributions.  I would like to either
> 
> 1) determine the mean and SD for each of the two distributions
> 
> OR
> 
> 2) get some (bayesian ?) statistic that estimates how likely an observation
is to belong to the left-hand or right-hand distribution
> 
> In case I'm using the wrong language, my data looks something like
this:
> 
> B <- rnorm(500,40,10)
> H <- rnorm(500,80,5 )
> N <- runif(200,0,99)
> D <- c(B,H,N)
> 
> Where B=background, H=hits, N=noise, and D=my observed distribution
> 
> I have seen analyses like this in the past, but I can't remember what
it is called.  If somebody out there can point me towards an R function, or even
the cannonical name for this kind of model, I think I can write the necessary
code.
> 
> Thanks in advance,
> Mark
> 
> ______________________________________________
> 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.
> 
-- 


**************************
Thomas Cornulier

Mammal Research Institute
Polish Academy of Sciences
ul. Waszkiewicza 1c
17-230 Bialowieza
http://www.zbs.bialowieza.pl/?en

tel (0048) 85 682 77 88

<mmiller at mail.usp.edu> wrote in
news:BAY105-W146A57D6AFC49F8003C73E8F210 at phx.gbl: 
> 
> Recently, I have been working with some data that look like two
> overlapping gaussian distributions.  I would like to either 
> 
> 1) determine the mean and SD for each of the two distributions
> 
> OR
> 
> 2) get some (bayesian ?) statistic that estimates how likely an
> observation is to belong to the left-hand or right-hand distribution 
> 
> In case I'm using the wrong language, my data looks something like
> this: 
> 
> B <- rnorm(500,40,10)
> H <- rnorm(500,80,5 )
> N <- runif(200,0,99)
> D <- c(B,H,N)
First hit on <http://www.rseek.org/> with: functions mixture gaussians

<http://research.stowers-institute.org/efg/R/Statistics/MixturesOfDistributions/index.htm>

Task Views:
<http://cran.us.r-project.org/web/views/Cluster.html>
<http://cran.r-project.org/web/views/Multivariate.html>
<http://cran.r-project.org/web/views/Bayesian.html>

nor1mix: Normal (1-d) Mixture Models (S3 Classes and Methods)
<http://lib.stat.cmu.edu/R/CRAN/>  

In the nor1mix documentation, M?chler recommends the mclust package 
for estimation.

-- 
David Winsemius

it could be a finite mixture.
take a look at flexmix package.

On Feb 18, 2008 11:03 AM,  <mmiller at mail.usp.edu>
wrote:
>
> Recently, I have been working with some data that look like two overlapping
gaussian distributions.  I would like to either
>
> 1) determine the mean and SD for each of the two distributions
>
> OR
>
> 2) get some (bayesian ?) statistic that estimates how likely an observation
is to belong to the left-hand or right-hand distribution
>
> In case I'm using the wrong language, my data looks something like
this:
>
> B <- rnorm(500,40,10)
> H <- rnorm(500,80,5 )
> N <- runif(200,0,99)
> D <- c(B,H,N)
>
> Where B=background, H=hits, N=noise, and D=my observed distribution
>
> I have seen analyses like this in the past, but I can't remember what
it is called.  If somebody out there can point me towards an R function, or even
the cannonical name for this kind of model, I think I can write the necessary
code.
>
> Thanks in advance,
> Mark
>
> ______________________________________________
> 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.
>


-- 
==============================WenSui Liu
ChoicePoint Precision Marketing
Phone: 678-893-9457
Email : wensui.liu at choicepoint.com
Blog   : statcompute.spaces.live.com

take a look at:


Du, 2002, Master Thesis, http://www.math.mcmaster.ca/peter/mix/Rmix.pdf

Macdonald, P., 2003, RMIX routine for R,
http://www.math.mcmaster.ca/peter/mix/mix.html

I don't think this package was actually posted on CRAN (the mix package on
CRAN is a different one as far as i remember) - maybe because Macdonald has also
a commercial package - or maybe had .... but you can model your distribution
with 2 or maybe more normal or log-normal distributions - as you see fit - and
does an ANOVA test to see if your modeling is statistically significant or not.
You will get also mean and standard distribution for each of your modeling
distributions.

Hope this helps,

Monica

______________________________________________________________
Message: 41
Date: Mon, 18 Feb 2008 16:03:10 +0000
From: 
Subject: [R] newbie (me) needs to model distribution as two
overlapping gaussians
To: 
Message-ID: 
Content-Type: text/plain; charset="iso-8859-1"


Recently, I have been working with some data that look like two overlapping
gaussian distributions. I would like to either

1) determine the mean and SD for each of the two distributions

OR

2) get some (bayesian ?) statistic that estimates how likely an observation is
to belong to the left-hand or right-hand distribution

In case I'm using the wrong language, my data looks something like this:

B <- rnorm(500,40,10)
H <- rnorm(500,80,5 )
N <- runif(200,0,99)
D <- c(B,H,N)

Where B=background, H=hits, N=noise, and D=my observed distribution

I have seen analyses like this in the past, but I can't remember what it is
called. If somebody out there can point me towards an R function, or even the
cannonical name for this kind of model, I think I can write the necessary code.

Thanks in advance,
Mark



_________________________________________________________________
[[elided Hotmail spam]]