[R] Off Topic: Re: Calculating NOEL using R and logistic regression - Toxicology

[Bert Gunter]

Below.
-- Bert

On Tue, Apr 3, 2012 at 1:47 PM, Danielle Duncan <dlduncan2 at alaska.edu> wrote:
> Thanks for the response, I should have clarified that the NOEL is the
> smallest dose above which there is a statistically significant effect.
>
This is not a scientifically meaningful nor defensible definition as
it is stochastic, depends on the test used, design, level chosen, etc.

-- Bert


> On Tue, Apr 3, 2012 at 12:44 PM, Drew Tyre <atyre2 at unl.edu> wrote:
>
>> Is this the smallest observed dose that has an effect? If so, then you
>> don't need the glm to find it. Here is a simulated example:
>>
>> set.seed(101)
>> X = rep(1:10,each=10)
>> lp = -5 + 0.5*X
>> Y = rbinom(length(X),size=1,p=1/(1+exp(-lp)))
>> # is this the NOEL?
>> min(X[Y==1]) # picks out observations with adverse effects, chooses the
>> smallest value
>> glmfit = glm(Y~X,family=binomial)
>> plot(1:10, predict(glmfit,newdata=data.frame(X=1:10),type="response"),
>> type="l",ylim=c(0,1),xlab="X",ylab="Y")
>> rug(jitter(X[Y==0]),side=1)
>> rug(jitter(X[Y==1]),side=3)
>>
>> On Tue, Apr 3, 2012 at 3:19 PM, Danielle Duncan <dlduncan2 at alaska.edu>wrote:
>>
>>> Thanks, that is interesting, but what I'm really after is an easy "no
>>> observed effect level", using a binomial logistic model ie glm. Have a
>>> great day!
>>>
>>> On Mon, Apr 2, 2012 at 3:38 PM, vito.muggeo <vito.muggeo at unipa.it> wrote:
>>>
>>> > dear Danielle,
>>> >
>>> > The NOEL is a threshold value or breakpoint in the range of dose. Have a
>>> > look to the
>>> > package segmented to estimate a GLM with unknown breakpoints. The code
>>> > (untested) should
>>> > be something like
>>> >
>>> > library(segmented)
>>> > o<-glm(y~1, family=binomial)
>>> > os<-segmented(o, ~dose, psi=starting_psi)
>>> >
>>> > Also the package segmented includes the dataset down that can be useful
>>> as
>>> > an example..
>>> >
>>> > data(down)
>>> > with(down, plot(age, cases/births))
>>> >
>>> > There is a paper of mine on R news 2008 discussing the package..
>>> >
>>> > hope this helps you,
>>> > vito
>>> >
>>> >
>>> >
>>> > On Mon, 2 Apr 2012 14:45:06 -0800, Danielle Duncan wrote
>>> > > Hello, I used the glm function in R to fit a dose-response
>>> relationship
>>> > and
>>> > > then have been using dose.p to calculate the LC50, however I would
>>> like
>>> > to
>>> > > calculate the NOEL (no observed effect level), ie the lowest dose
>>> above
>>> > > which responses start occurring. Does anyone know how to do this?
>>> > >
>>> > >       [[alternative HTML version deleted]]
>>> > >
>>> > > ______________________________________________
>>> > > 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.
>>> >
>>> >
>>> > --
>>> > Open WebMail Project (http://openwebmail.org)
>>> >
>>> >
>>>
>>>        [[alternative HTML version deleted]]
>>>
>>> ______________________________________________
>>> 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.
>>>
>>
>>
>>
>> --
>> Drew Tyre
>>
>> School of Natural Resources
>> University of Nebraska-Lincoln
>> 416 Hardin Hall, East Campus
>> 3310 Holdrege Street
>> Lincoln, NE 68583-0974
>>
>> phone: +1 402 472 4054
>> fax: +1 402 472 2946
>> email: atyre2 at unl.edu
>> http://snr.unl.edu/tyre
>> http://aminpractice.blogspot.com
>> http://www.flickr.com/photos/atiretoo
>>
>
>        [[alternative HTML version deleted]]
>
> ______________________________________________
> 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.



-- 

Bert Gunter
Genentech Nonclinical Biostatistics

Internal Contact Info:
Phone: 467-7374
Website:
http://pharmadevelopment.roche.com/index/pdb/pdb-functional-groups/pdb-biostatistics/pdb-ncb-home.htm