[R-sig-ME] question about a GAM model (dani)

dani orchidn @ending from live@com
Wed Aug 29 22:08:27 CEST 2018 

Hello Dr. Zuur,


Thank you so much for your prompt and detailed response. That is very helpful! Thank so much for your advice!


I also have another issue that is not clear to me and I could not find any information about that so far. Assuming that my model includes many parametric covariates, does it make any sense to standardize coefficients in a binomial GAM model and report both unstandardized and standardized coefficients for the parametric coefficients in a manuscript? I have never seen that in the literature, so I really do not know how to approach this issue.


Best regards,

Dani




________________________________
From: Highland Statistics Ltd <highstat using highstat.com>
Sent: Wednesday, August 29, 2018 12:55 PM
To: dani; r-sig-mixed-models using r-project.org
Subject: Re: [R-sig-ME] question about a GAM model (dani)



On 29/08/2018 20:42, dani wrote:

Hello Dr. Zuur,


Thank you so much for your message!


Dani,

I am only using this model for educational purposes, I am just playing with a dataset of 500 observations. Variables x1 and x2 are covariates and they are both displaying non-parametric associations with the outcome. The x3 variable is the variable of interest.


The fact that x1 vs Y and x2 vs Y show non-linear patterns is no 100% guarantee that in a model with Y ~ X1 + X2 each of them also show a non-linear pattern.


I noticed the value of 1 for edfs for the covariate and for the variable of interest so I asked myself if I should not remove the parametric term and re-run the model is situations like these.

That is a sensible line of thinking. The AIC is also your friend here.


If this happens when I conduct an analysis for a study, do I present such results or I re-run the model without smoothers on x2 and x3, even though in bivariate associations with the outcome, x2 and x3 showed non-parametric associations.

My strategy for GAMs is to only use those covariates as smoothers that make (biological) sense. You can then either start with a parametric model (e.g. a GLM) and inspect residuals, or start with a (simple) GAM and see what the edf tells you (or see how the smoothers look like) and potentially move back to a GLM (but note that link functions can also cause non-linear patterns, or remove non-linear patterns). This is the chicken or the egg.

Alain





Thank you so much for your suggestions, I will definitely look at the two books again, they are always useful!

Best,

Dani


________________________________
From: R-sig-mixed-models <r-sig-mixed-models-bounces using r-project.org><mailto:r-sig-mixed-models-bounces using r-project.org> on behalf of Highland Statistics Ltd <highstat using highstat.com><mailto:highstat using highstat.com>
Sent: Wednesday, August 29, 2018 12:30 PM
To: r-sig-mixed-models using r-project.org<mailto:r-sig-mixed-models using r-project.org>
Subject: Re: [R-sig-ME] question about a GAM model (dani)




------------------------------

Message: 4
Date: Wed, 29 Aug 2018 17:38:44 +0000
From: dani <orchidn using live.com><mailto:orchidn using live.com>
To: Ben Bolker <bbolker using gmail.com><mailto:bbolker using gmail.com>, "r-sig-mixed-models using r-project.org"<mailto:r-sig-mixed-models using r-project.org>
        <r-sig-mixed-models using r-project.org><mailto:r-sig-mixed-models using r-project.org>
Subject: Re: [R-sig-ME] question about a GAM model
Message-ID:
        <BYAPR06MB38321CD8470AE34E57716661D6090 using BYAPR06MB3832.namprd06.prod.outlook.com><mailto:BYAPR06MB38321CD8470AE34E57716661D6090 using BYAPR06MB3832.namprd06.prod.outlook.com>

Content-Type: text/plain; charset="utf-8"

Thank you very much Udita and Dr. Bolker for your responses.

It is still not clear to me how should I proceed. Would anyone else be able help with this issue,?




Dani,

GAMs are useful if you use them with care, but confusing if you just apply them because someone else is doing it as well.
Perhaps you should first ask yourself the question why you are applying a GAM. Then focus on the question
whether the output makes sense.

Based on your output it seems that nothing is important (not even as parametric terms). But I am not familiar with your
data; things like collinearity can mess up the shape of smoothers. And you don't mention the size of your data set neither.

I suggest that you have a go at Wood (2017), or if I may be bold enough to self-cite....try our Beginner's Guide to GAM (2012).



Kind regards,

Alain Zuur






  Best regards,

Dani


Sent from Outlook<http://aka.ms/weboutlook>


________________________________
From: R-sig-mixed-models <r-sig-mixed-models-bounces using r-project.org><mailto:r-sig-mixed-models-bounces using r-project.org> on behalf of Ben Bolker <bbolker using gmail.com><mailto:bbolker using gmail.com>
Sent: Tuesday, August 28, 2018 6:19 AM
To: r-sig-mixed-models using r-project.org<mailto:r-sig-mixed-models using r-project.org>
Subject: Re: [R-sig-ME] question about a GAM model



    Don't forget to run k.check() on your model to see if you specified a
large enough basis dimension  to start with ...

On 2018-08-28 05:51 AM, Bansal, Udita wrote:
> Hi Dani,
>
> I don�t know much about GAM but I know you can look at the plots for fitted model results to check if there is any curvature. You can use the following code:
>
> par(mfrow = c(1,3))
> plot(GAMmodel)
>
> Bests
> Udita
>
> On 28/08/18, 1:58 AM, "R-sig-mixed-models on behalf of dani" <r-sig-mixed-models-bounces using r-project.org on behalf of orchidn using live.com><mailto:r-sig-mixed-models-bounces using r-project.orgonbehalfoforchidn@live.com> wrote:
>
>     Hi everyone,
>
>
>     I have a question about a GAM model where I included three non-parametric terms. I obtained the results below. can I conclude that the associations were in fact linear and run a final GLM model without including splines? To me it seems unnecessary to include splines in the final model. How should I report these results?
>
>
>     # Approximate significance of smooth terms:
>     #                 edf Ref.df Chi.sq p-value
>     # s(x1)      1.61   2.01   1.17   0.550
>     # s(x2)      1.00   1.00   0.00   0.955
>     # s(x3)      1.00   1.00   4.61   0.032 *
>
>     Thank you very much,
>     Dani
>
>
>
>
>
>        [[alternative HTML version deleted]]
>
>     _______________________________________________
>     R-sig-mixed-models using r-project.org<mailto:R-sig-mixed-models using r-project.org> mailing list
>     https://stat.ethz.ch/mailman/listinfo/r-sig-mixed-models
>
>
> _______________________________________________
> R-sig-mixed-models using r-project.org<mailto:R-sig-mixed-models using r-project.org> mailing list
> https://stat.ethz.ch/mailman/listinfo/r-sig-mixed-models
>

_______________________________________________
R-sig-mixed-models using r-project.org<mailto:R-sig-mixed-models using r-project.org> mailing list
https://stat.ethz.ch/mailman/listinfo/r-sig-mixed-models

        [[alternative HTML version deleted]]




------------------------------

Subject: Digest Footer

_______________________________________________
R-sig-mixed-models mailing list
R-sig-mixed-models using r-project.org<mailto:R-sig-mixed-models using r-project.org>
https://stat.ethz.ch/mailman/listinfo/r-sig-mixed-models


------------------------------

End of R-sig-mixed-models Digest, Vol 140, Issue 34
***************************************************

--

Dr. Alain F. Zuur
Highland Statistics Ltd.
9 St Clair Wynd
AB41 6DZ Newburgh, UK
Email: highstat using highstat.com<mailto:highstat using highstat.com>
URL:   www.highstat.com<http://www.highstat.com>

And:
NIOZ Royal Netherlands Institute for Sea Research,
Department of Coastal Systems, and Utrecht University,
P.O. Box 59, 1790 AB Den Burg,
Texel, The Netherlands



Author of:
1. Beginner's Guide to Spatial, Temporal and Spatial-Temporal Ecological Data Analysis with R-INLA. (2017).
2. Beginner's Guide to Zero-Inflated Models with R (2016).
3. Beginner's Guide to Data Exploration and Visualisation with R (2015).
4. Beginner's Guide to GAMM with R (2014).
5. Beginner's Guide to GLM and GLMM with R (2013).
6. Beginner's Guide to GAM with R (2012).
7. Zero Inflated Models and GLMM with R (2012).
8. A Beginner's Guide to R (2009).
9. Mixed effects models and extensions in ecology with R (2009).
10. Analysing Ecological Data (2007).

_______________________________________________
R-sig-mixed-models using r-project.org<mailto:R-sig-mixed-models using r-project.org> mailing list
https://stat.ethz.ch/mailman/listinfo/r-sig-mixed-models


--

Dr. Alain F. Zuur
Highland Statistics Ltd.
9 St Clair Wynd
AB41 6DZ Newburgh, UK
Email: highstat using highstat.com<mailto:highstat using highstat.com>
URL:   www.highstat.com<http://www.highstat.com>

And:
NIOZ Royal Netherlands Institute for Sea Research,
Department of Coastal Systems, and Utrecht University,
P.O. Box 59, 1790 AB Den Burg,
Texel, The Netherlands



Author of:
1. Beginner's Guide to Spatial, Temporal and Spatial-Temporal Ecological Data Analysis with R-INLA. (2017).
2. Beginner's Guide to Zero-Inflated Models with R (2016).
3. Beginner's Guide to Data Exploration and Visualisation with R (2015).
4. Beginner's Guide to GAMM with R (2014).
5. Beginner's Guide to GLM and GLMM with R (2013).
6. Beginner's Guide to GAM with R (2012).
7. Zero Inflated Models and GLMM with R (2012).
8. A Beginner's Guide to R (2009).
9. Mixed effects models and extensions in ecology with R (2009).
10. Analysing Ecological Data (2007).



	[[alternative HTML version deleted]]

More information about the R-sig-mixed-models mailing list