Help MARS Functions

classic Classic list List threaded Threaded
4 messages Options
Reply | Threaded
Open this post in threaded view
|

Help MARS Functions

Miguel Cañedo
Hi!

I am using the mars.csv function to predict the occurrences of diatoms in
the Ebro basin (Spain) and I am finding some problems that I can´t solve. I
get the following error:

Creating predictions for subsets...
glm.fit: fitted probabilities numerically 0 or 1 occurred

I am attaching the predictors + species matrix (DiatomsCV.csv), the
deviance resulting from applying the mars.glm function, the MARS functions
and the R code.

Can anyone take a quick look at it? I am really stuck...

Thank you very much in advance!

--
*“All things that are still to come lie in uncertainty; live straightway!” *

*― Seneca <https://www.goodreads.com/author/show/4918776.Seneca>, On the
Shortness of Life <https://www.goodreads.com/work/quotes/1374471>*

*Miguel Cañedo-Argüelles*
https://www.researchgate.net/profile/Miguel_Canedo-Argueelles/
http://www.linkedin.com/pub/miguel-cañedo-argüelles/23/4aa/249

*BETA Group, University of Vic*
http://www.uvic.cat/es/centro-tecnol%C3%B3gico-beta
<http://www.uvic.es/en/showrecerca/94>


*FEM Group, University of Barcelona*http://www.ub.edu/fem/index.php
<http://www.ub.edu/fem/>
_______________________________________________
R-sig-ecology mailing list
[hidden email]
https://stat.ethz.ch/mailman/listinfo/r-sig-ecology
Reply | Threaded
Open this post in threaded view
|

Re: Help MARS Functions

Evan Palmer-Young-2
It looks like you have complete separation, meaning that some groups either
have occurrences at every observation (p=1) or for no observations (p=0).

Maybe you could combine some sites or treatments with low sample size

Fyi attachments get scrubbed from the list. A link to a cloud file might
work better

Good luck!


On Feb 16, 2017 03:26, "Miguel Cañedo" <[hidden email]> wrote:

Hi!

I am using the mars.csv function to predict the occurrences of diatoms in
the Ebro basin (Spain) and I am finding some problems that I can´t solve. I
get the following error:

Creating predictions for subsets...
glm.fit: fitted probabilities numerically 0 or 1 occurred

I am attaching the predictors + species matrix (DiatomsCV.csv), the
deviance resulting from applying the mars.glm function, the MARS functions
and the R code.

Can anyone take a quick look at it? I am really stuck...

Thank you very much in advance!

--
*“All things that are still to come lie in uncertainty; live straightway!” *

*― Seneca <https://www.goodreads.com/author/show/4918776.Seneca>, On the
Shortness of Life <https://www.goodreads.com/work/quotes/1374471>*

*Miguel Cañedo-Argüelles*
https://www.researchgate.net/profile/Miguel_Canedo-Argueelles/
http://www.linkedin.com/pub/miguel-cañedo-argüelles/23/4aa/249

*BETA Group, University of Vic*
http://www.uvic.cat/es/centro-tecnol%C3%B3gico-beta
<http://www.uvic.es/en/showrecerca/94>


*FEM Group, University of Barcelona*http://www.ub.edu/fem/index.php
<http://www.ub.edu/fem/>
_______________________________________________
R-sig-ecology mailing list
[hidden email]
https://stat.ethz.ch/mailman/listinfo/r-sig-ecology

        [[alternative HTML version deleted]]

_______________________________________________
R-sig-ecology mailing list
[hidden email]
https://stat.ethz.ch/mailman/listinfo/r-sig-ecology
Reply | Threaded
Open this post in threaded view
|

Re: Help MARS Functions

Miguel Cañedo
Hi Evan!

Thank you so much! I had already removed species with less than 5
occurrences. I even tried by only leaving those species with good model
performance. Yet, I get the same error.

I have uploaded the files here (thanks for the advice!):

https://drive.google.com/open?id=0B0sFWHdTbJiLeFhfdUlfSUJOUUE

If anyone could have a quick look at the R code I would be really grateful.

Best, Miguel.

2017-02-16 12:16 GMT+01:00 Evan Palmer-Young <[hidden email]>:

> It looks like you have complete separation, meaning that some groups
> either have occurrences at every observation (p=1) or for no observations
> (p=0).
>
> Maybe you could combine some sites or treatments with low sample size
>
> Fyi attachments get scrubbed from the list. A link to a cloud file might
> work better
>
> Good luck!
>
>
> On Feb 16, 2017 03:26, "Miguel Cañedo" <[hidden email]> wrote:
>
> Hi!
>
> I am using the mars.csv function to predict the occurrences of diatoms in
> the Ebro basin (Spain) and I am finding some problems that I can´t solve. I
> get the following error:
>
> Creating predictions for subsets...
> glm.fit: fitted probabilities numerically 0 or 1 occurred
>
> I am attaching the predictors + species matrix (DiatomsCV.csv), the
> deviance resulting from applying the mars.glm function, the MARS functions
> and the R code.
>
> Can anyone take a quick look at it? I am really stuck...
>
> Thank you very much in advance!
>
> --
> *“All things that are still to come lie in uncertainty; live straightway!”
> *
>
> *― Seneca <https://www.goodreads.com/author/show/4918776.Seneca>, On the
> Shortness of Life <https://www.goodreads.com/work/quotes/1374471>*
>
> *Miguel Cañedo-Argüelles*
> https://www.researchgate.net/profile/Miguel_Canedo-Argueelles/
> http://www.linkedin.com/pub/miguel-cañedo-argüelles/23/4aa/249
>
> *BETA Group, University of Vic*
> http://www.uvic.cat/es/centro-tecnol%C3%B3gico-beta
> <http://www.uvic.es/en/showrecerca/94>
>
>
> *FEM Group, University of Barcelona*http://www.ub.edu/fem/index.php
> <http://www.ub.edu/fem/>
> _______________________________________________
> R-sig-ecology mailing list
> [hidden email]
> https://stat.ethz.ch/mailman/listinfo/r-sig-ecology
>
>
>


--
*“All things that are still to come lie in uncertainty; live straightway!” *

*― Seneca <https://www.goodreads.com/author/show/4918776.Seneca>, On the
Shortness of Life <https://www.goodreads.com/work/quotes/1374471>*

*Miguel Cañedo-Argüelles*
https://www.researchgate.net/profile/Miguel_Canedo-Argueelles/
http://www.linkedin.com/pub/miguel-cañedo-argüelles/23/4aa/249

*BETA Group, University of Vic*
http://www.uvic.cat/es/centro-tecnol%C3%B3gico-beta
<http://www.uvic.es/en/showrecerca/94>


*FEM Group, University of Barcelona*http://www.ub.edu/fem/index.php
<http://www.ub.edu/fem/>

        [[alternative HTML version deleted]]

_______________________________________________
R-sig-ecology mailing list
[hidden email]
https://stat.ethz.ch/mailman/listinfo/r-sig-ecology
Reply | Threaded
Open this post in threaded view
|

Re: Help MARS Functions

Evan Palmer-Young-2
Lmer also has a predict() function. Or you can use lsmeans to get fitted
values and se or ci for particular factor or covariate combos.

Boxplot the data first to check for zeroes and also ones though
Good luck!

On Feb 16, 2017 06:43, "Miguel Cañedo" <[hidden email]> wrote:

> Hi Evan!
>
> Thank you so much! I had already removed species with less than 5
> occurrences. I even tried by only leaving those species with good model
> performance. Yet, I get the same error.
>
> I have uploaded the files here (thanks for the advice!):
>
> https://drive.google.com/open?id=0B0sFWHdTbJiLeFhfdUlfSUJOUUE
>
> If anyone could have a quick look at the R code I would be really grateful.
>
> Best, Miguel.
>
> 2017-02-16 12:16 GMT+01:00 Evan Palmer-Young <[hidden email]>:
>
>> It looks like you have complete separation, meaning that some groups
>> either have occurrences at every observation (p=1) or for no observations
>> (p=0).
>>
>> Maybe you could combine some sites or treatments with low sample size
>>
>> Fyi attachments get scrubbed from the list. A link to a cloud file might
>> work better
>>
>> Good luck!
>>
>>
>> On Feb 16, 2017 03:26, "Miguel Cañedo" <[hidden email]> wrote:
>>
>> Hi!
>>
>> I am using the mars.csv function to predict the occurrences of diatoms in
>> the Ebro basin (Spain) and I am finding some problems that I can´t solve.
>> I
>> get the following error:
>>
>> Creating predictions for subsets...
>> glm.fit: fitted probabilities numerically 0 or 1 occurred
>>
>> I am attaching the predictors + species matrix (DiatomsCV.csv), the
>> deviance resulting from applying the mars.glm function, the MARS functions
>> and the R code.
>>
>> Can anyone take a quick look at it? I am really stuck...
>>
>> Thank you very much in advance!
>>
>> --
>> *“All things that are still to come lie in uncertainty; live
>> straightway!” *
>>
>> *― Seneca <https://www.goodreads.com/author/show/4918776.Seneca>, On the
>> Shortness of Life <https://www.goodreads.com/work/quotes/1374471>*
>>
>> *Miguel Cañedo-Argüelles*
>> https://www.researchgate.net/profile/Miguel_Canedo-Argueelles/
>> http://www.linkedin.com/pub/miguel-cañedo-argüelles/23/4aa/249
>>
>> *BETA Group, University of Vic*
>> http://www.uvic.cat/es/centro-tecnol%C3%B3gico-beta
>> <http://www.uvic.es/en/showrecerca/94>
>>
>>
>> *FEM Group, University of Barcelona*http://www.ub.edu/fem/index.php
>> <http://www.ub.edu/fem/>
>> _______________________________________________
>> R-sig-ecology mailing list
>> [hidden email]
>> https://stat.ethz.ch/mailman/listinfo/r-sig-ecology
>>
>>
>>
>
>
> --
> *“All things that are still to come lie in uncertainty; live
> straightway!” *
>
> *― Seneca <https://www.goodreads.com/author/show/4918776.Seneca>, On the
> Shortness of Life <https://www.goodreads.com/work/quotes/1374471>*
>
> *Miguel Cañedo-Argüelles*
> https://www.researchgate.net/profile/Miguel_Canedo-Argueelles/
> http://www.linkedin.com/pub/miguel-cañedo-argüelles/23/4aa/249
>
> *BETA Group, University of Vic*
> http://www.uvic.cat/es/centro-tecnol%C3%B3gico-beta
> <http://www.uvic.es/en/showrecerca/94>
>
>
> *FEM Group, University of Barcelona*http://www.ub.edu/fem/index.php
> <http://www.ub.edu/fem/>
>

        [[alternative HTML version deleted]]

_______________________________________________
R-sig-ecology mailing list
[hidden email]
https://stat.ethz.ch/mailman/listinfo/r-sig-ecology