[R-sig-ME] glmmadmb beta - error

[Highland Statistics Ltd]

Ludovico...it would be nice if you can cc the mixed modelling mailing list.

Whatever you do, you need to get more data. The model below is applying 
a model with 5 regression parameters on 25 observations. And you do it 
twice. And you may still need a beta distribution.

What you can do is fit your original beta (or Gaussian) mixed model and 
do some simulations from the model, and see whether you are able to 
estimate similar parameter estimates from the simulated data. If you are 
lucky and you have strong patterns in your data, than perhaps you may 
get away with your low sample size.


I understand that for some studies 50 observations represent a lot of 
work/time/money, but what is the point in spending all that time if at 
the end of the day the sample size is too low to answer your questions?

Alain




On 23/02/2016 15:42, Ludovico Frate wrote:
> Thank you!
> Since I have repeated measure for each plot (two observation for 
> plot), could be an idea subtracting the cover values in pairs (i.e. 
> plot 1 cover 2015-cover 2001, plot 2 cover 2015-cover 2001, and so on) 
> and then try to fit a simple linear model with the new dependent 
> variable?
>
> lm(cover_delta~1 + direction+time)
>
> Regards,
> Ludovico
>
> ------------------------------------------------------------------------
> Subject: Re: [R-sig-ME] glmmadmb beta - error
> To: ludovicofrate at hotmail.it; r-sig-mixed-models at r-project.org
> From: highstat at highstat.com
> Date: Tue, 23 Feb 2016 12:49:51 +0000
>
> The fact that the software runs properly, or not, is irrelevant. You 
> need to have at least 10 - 15 - 20 observations per regression 
> parameter for an ordinary regression model. Using only 50 observations 
> for the beta mixed model is not the best thing to do. Maybe this is 
> not what you wanted to hear.
>
> Alain
>
> On 23/02/2016 12:43, Ludovico Frate wrote:
>
>     Thanks for your reply!
>     I successfully fitted another model with the same number of
>     observation (and structure) and the only difference is that cover
>     values were more centered on the mean values.
>
>     I just tried to fit the same model by adding a constant cover
>     value to all observations (+0.20) and the model works
>
>     I think that the issue derives from very small cover values...
>
>     summary(cover_lf$cover) (divided by 100)
>
>     Min. 1st Qu.  Median    Mean 3rd Qu.    Max.
>     0.0000  0.0050  0.0130  0.0252  0.0385  0.1300
>
>     Call:
>     glmmadmb(formula = cover ~ time + summit + direction + (1 | plot_id),
>         data = cover_lf, family = "beta")
>
>     AIC: -228.7
>
>     Coefficients:
>                 Estimate Std. Error z value Pr(>|z|)
>     (Intercept)  -1.2241     0.0476  -25.72 <2e-16 ***
>     time2015     -0.0688     0.0337   -2.04   0.0410 *
>     summitMAC    -0.1407     0.0437   -3.22   0.0013 **
>     directionN    0.1028     0.0574    1.79   0.0733 .
>     directionS    0.0760     0.0624    1.22   0.2238
>     directionW    0.0986     0.0574    1.72   0.0860 .
>     ---
>     Signif. codes:  0 ‘***’ 0.001 ‘**’ 0.01 ‘*’ 0.05 ‘.’ 0.1 ‘ ’ 1
>
>     Number of observations: total=50, plot_id=25
>     Random effect variance(s):
>     Group=plot_id
>                 Variance  StdDev
>     (Intercept) 0.003394 0.05826
>
>     Beta dispersion parameter: 403.43 (std. err.: 0.00012419)
>
>     Log-likelihood: 122.34
>
>
> -- 
> Dr. Alain F. Zuur
>
> First author of:
> 1. Beginner's Guide to GAMM with R (2014).
> 2. Beginner's Guide to GLM and GLMM with R (2013).
> 3. Beginner's Guide to GAM with R (2012).
> 4. Zero Inflated Models and GLMM with R (2012).
> 5. A Beginner's Guide to R (2009).
> 6. Mixed effects models and extensions in ecology with R (2009).
> 7. Analysing Ecological Data (2007).
>
> Highland Statistics Ltd.
> 9 St Clair Wynd
> UK - AB41 6DZ Newburgh
> Tel:   0044 1358 788177
> Email:highstat at highstat.com <mailto:highstat at highstat.com>
> URL:www.highstat.com <http://www.highstat.com>

-- 
Dr. Alain F. Zuur

First author of:
1. Beginner's Guide to GAMM with R (2014).
2. Beginner's Guide to GLM and GLMM with R (2013).
3. Beginner's Guide to GAM with R (2012).
4. Zero Inflated Models and GLMM with R (2012).
5. A Beginner's Guide to R (2009).
6. Mixed effects models and extensions in ecology with R (2009).
7. Analysing Ecological Data (2007).

Highland Statistics Ltd.
9 St Clair Wynd
UK - AB41 6DZ Newburgh
Tel:   0044 1358 788177
Email: highstat at highstat.com
URL:   www.highstat.com


	[[alternative HTML version deleted]]