interaction terms interpretation

Hi all, I?m running a mixed model with poisson error distribution as I have
count data of a shorebird species. My predictors are AMGP (another
shorebird that is usually associated to "my" species and probably has
ecological influence on it), grass height and field distance to the lagoon.

I?m having a hard time trying to interpret second order interactions. I
appreciate your comments of references recommendations. Here is the output
of my model. All the best, Joaqu?n.

Call:glmmadmb(formula = BBSA ~ AMGP * Grass_height *
Distance_to_lagoon +     (1 | Field_name) + offset(log(Field_area.o)),
data = my4S,     family = "nbinom1", zeroInflation = TRUE)
AIC: 226
Coefficients:                                     Estimate Std. Error
z value Pr(>|z|)    (Intercept)                           -5.8797
1.2250   -4.80  1.6e-06 ***AMGP
3.3086     1.0246    3.23   0.0012 ** Grass_height
     -4.1592     1.9961   -2.08   0.0372 *  Distance_to_lagoon
           -0.5029     1.3929   -0.36   0.7180    AMGP:Grass_height
                  2.9442     1.5142    1.94   0.0518 .
AMGP:Distance_to_lagoon               -2.4695     0.9632   -2.56
0.0104 *  Grass_height:Distance_to_lagoon        0.0339     2.0065
0.02   0.9865    AMGP:Grass_height:Distance_to_lagoon  -4.0363
1.3264   -3.04   0.0023 ** ---Signif. codes:  0 ?***? 0.001 ?**? 0.01
?*? 0.05 ?.? 0.1 ? ? 1
Number of observations: total=64, Field_name=16 Random effect
variance(s):Group=Field_name            Variance StdDev(Intercept)
0.03721 0.1929
Negative binomial dispersion parameter: 20.633 (std. err.:
7.4042)Zero-inflation: 1e-06  (std. err.:  6.8336e-09 )
Log-likelihood: -102


--
*Joaqu?n Aldabe*

*Grupo Biodiversidad, Ambiente y Sociedad*
Centro Universitario de la Regi?n Este, Universidad de la Rep?blica
Ruta 15 (y Ruta 9), Km 28.500, Departamento de Rocha

*Departamento de Conservaci?n*
Aves Uruguay
BirdLife International
Canelones 1164, Montevideo

https://sites.google.com/site/joaquin.aldabe
<https://sites.google.com/site/perfilprofesionaljoaquinaldabe>

        [[alternative HTML version deleted]]

Dear Joaquin,

Your mail is mangled because you send it in HTML. This makes the model
output very hard to read.

It's impossible to interpret model output without thorough knowledge of the
dataset. I suggest that you try to make some graphs of the model.

Best regards,

ir. Thierry Onkelinx
Instituut voor natuur- en bosonderzoek / Research Institute for Nature and
Forest
team Biometrie & Kwaliteitszorg / team Biometrics & Quality Assurance
Kliniekstraat 25
1070 Anderlecht
Belgium

To call in the statistician after the experiment is done may be no more
than asking him to perform a post-mortem examination: he may be able to say
what the experiment died of. ~ Sir Ronald Aylmer Fisher
The plural of anecdote is not data. ~ Roger Brinner
The combination of some data and an aching desire for an answer does not
ensure that a reasonable answer can be extracted from a given body of data.
~ John Tukey

2015-08-07 21:27 GMT+02:00 Joaqu?n Aldabe <joaquin.aldabe at gmail.com>:

Hi all, I?m running a mixed model with poisson error distribution as I

have

count data of a shorebird species. My predictors are AMGP (another
shorebird that is usually associated to "my" species and probably has
ecological influence on it), grass height and field distance to the

lagoon.

I?m having a hard time trying to interpret second order interactions. I
appreciate your comments of references recommendations. Here is the

output

of my model. All the best, Joaqu?n.

Call:glmmadmb(formula = BBSA ~ AMGP * Grass_height *
Distance_to_lagoon +     (1 | Field_name) + offset(log(Field_area.o)),
data = my4S,     family = "nbinom1", zeroInflation = TRUE)
AIC: 226
Coefficients:                                     Estimate Std. Error
z value Pr(>|z|)    (Intercept)                           -5.8797
1.2250   -4.80  1.6e-06 ***AMGP
3.3086     1.0246    3.23   0.0012 ** Grass_height
     -4.1592     1.9961   -2.08   0.0372 *  Distance_to_lagoon
           -0.5029     1.3929   -0.36   0.7180    AMGP:Grass_height
                  2.9442     1.5142    1.94   0.0518 .
AMGP:Distance_to_lagoon               -2.4695     0.9632   -2.56
0.0104 *  Grass_height:Distance_to_lagoon        0.0339     2.0065
0.02   0.9865    AMGP:Grass_height:Distance_to_lagoon  -4.0363
1.3264   -3.04   0.0023 ** ---Signif. codes:  0 ?***? 0.001 ?**? 0.01
?*? 0.05 ?.? 0.1 ? ? 1
Number of observations: total=64, Field_name=16 Random effect
variance(s):Group=Field_name            Variance StdDev(Intercept)
0.03721 0.1929
Negative binomial dispersion parameter: 20.633 (std. err.:
7.4042)Zero-inflation: 1e-06  (std. err.:  6.8336e-09 )
Log-likelihood: -102


--
*Joaqu?n Aldabe*

*Grupo Biodiversidad, Ambiente y Sociedad*
Centro Universitario de la Regi?n Este, Universidad de la Rep?blica
Ruta 15 (y Ruta 9), Km 28.500, Departamento de Rocha

*Departamento de Conservaci?n*
Aves Uruguay
BirdLife International
Canelones 1164, Montevideo

https://sites.google.com/site/joaquin.aldabe
<https://sites.google.com/site/perfilprofesionaljoaquinaldabe>

        [[alternative HTML version deleted]]

        [[alternative HTML version deleted]]

There is actually a continuous third order interaction, which are just
about impossible to interpret.

As well as the graphs, which you always should start with, think about
what is sensible for the model. Should any of the predictors be log-scaled?
As AMGP is a count it wouldn't be surprising it it is, or use log(1+x) if
there are zeroes.

Look at diagnostic plots and make sure that everything is fine, especially
that there are no outliers which might be causing the interaction.

Ken

On 10 August 2015 at 17:02, Thierry Onkelinx <thierry.onkelinx at inbo.be>
wrote:

Dear Joaquin,

Your mail is mangled because you send it in HTML. This makes the model
output very hard to read.

It's impossible to interpret model output without thorough knowledge of
the
dataset. I suggest that you try to make some graphs of the model.

Best regards,

ir. Thierry Onkelinx
Instituut voor natuur- en bosonderzoek / Research Institute for Nature and
Forest
team Biometrie & Kwaliteitszorg / team Biometrics & Quality Assurance
Kliniekstraat 25
1070 Anderlecht
Belgium

To call in the statistician after the experiment is done may be no more
than asking him to perform a post-mortem examination: he may be able to
say
what the experiment died of. ~ Sir Ronald Aylmer Fisher
The plural of anecdote is not data. ~ Roger Brinner
The combination of some data and an aching desire for an answer does not
ensure that a reasonable answer can be extracted from a given body of
data.
~ John Tukey

2015-08-07 21:27 GMT+02:00 Joaqu?n Aldabe <joaquin.aldabe at gmail.com>:

Hi all, I?m running a mixed model with poisson error distribution as I

have

count data of a shorebird species. My predictors are AMGP (another
shorebird that is usually associated to "my" species and probably has
ecological influence on it), grass height and field distance to the

lagoon.

I?m having a hard time trying to interpret second order interactions. I
appreciate your comments of references recommendations. Here is the

output

of my model. All the best, Joaqu?n.

Call:glmmadmb(formula = BBSA ~ AMGP * Grass_height *
Distance_to_lagoon +     (1 | Field_name) + offset(log(Field_area.o)),
data = my4S,     family = "nbinom1", zeroInflation = TRUE)
AIC: 226
Coefficients:                                     Estimate Std. Error
z value Pr(>|z|)    (Intercept)                           -5.8797
1.2250   -4.80  1.6e-06 ***AMGP
3.3086     1.0246    3.23   0.0012 ** Grass_height
     -4.1592     1.9961   -2.08   0.0372 *  Distance_to_lagoon
           -0.5029     1.3929   -0.36   0.7180    AMGP:Grass_height
                  2.9442     1.5142    1.94   0.0518 .
AMGP:Distance_to_lagoon               -2.4695     0.9632   -2.56
0.0104 *  Grass_height:Distance_to_lagoon        0.0339     2.0065
0.02   0.9865    AMGP:Grass_height:Distance_to_lagoon  -4.0363
1.3264   -3.04   0.0023 ** ---Signif. codes:  0 ?***? 0.001 ?**? 0.01
?*? 0.05 ?.? 0.1 ? ? 1
Number of observations: total=64, Field_name=16 Random effect
variance(s):Group=Field_name            Variance StdDev(Intercept)
0.03721 0.1929
Negative binomial dispersion parameter: 20.633 (std. err.:
7.4042)Zero-inflation: 1e-06  (std. err.:  6.8336e-09 )
Log-likelihood: -102


--
*Joaqu?n Aldabe*

*Grupo Biodiversidad, Ambiente y Sociedad*
Centro Universitario de la Regi?n Este, Universidad de la Rep?blica
Ruta 15 (y Ruta 9), Km 28.500, Departamento de Rocha

*Departamento de Conservaci?n*
Aves Uruguay
BirdLife International
Canelones 1164, Montevideo

https://sites.google.com/site/joaquin.aldabe
<https://sites.google.com/site/perfilprofesionaljoaquinaldabe>

        [[alternative HTML version deleted]]

        [[alternative HTML version deleted]]

--

*Ken Beath*
Lecturer
Statistics Department
MACQUARIE UNIVERSITY NSW 2109, Australia

Phone: +61 (0)2 9850 8516

Level 2, AHH
http://stat.mq.edu.au/our_staff/staff_-_alphabetical/staff/beath,_ken/

CRICOS Provider No 00002J
This message is intended for the addressee named and m...{{dropped:12}}

Thankyoy very much Ken and Thierry. I'll try that. Is there any specific
function for making graphs of the model? I appreciate recommendations or
citations to explore how to graph model results. Thanks again. Joaquin
El 10/08/2015 07:36, "Ken Beath" <ken.beath at mq.edu.au> escribi?:

There is actually a continuous third order interaction, which are just
about impossible to interpret.

As well as the graphs, which you always should start with, think about
what is sensible for the model. Should any of the predictors be log-scaled?
As AMGP is a count it wouldn't be surprising it it is, or use log(1+x) if
there are zeroes.

Look at diagnostic plots and make sure that everything is fine, especially
that there are no outliers which might be causing the interaction.

Ken

On 10 August 2015 at 17:02, Thierry Onkelinx <thierry.onkelinx at inbo.be>
wrote:

Dear Joaquin,

Your mail is mangled because you send it in HTML. This makes the model
output very hard to read.

It's impossible to interpret model output without thorough knowledge of
the
dataset. I suggest that you try to make some graphs of the model.

Best regards,

ir. Thierry Onkelinx
Instituut voor natuur- en bosonderzoek / Research Institute for Nature and
Forest
team Biometrie & Kwaliteitszorg / team Biometrics & Quality Assurance
Kliniekstraat 25
1070 Anderlecht
Belgium

To call in the statistician after the experiment is done may be no more
than asking him to perform a post-mortem examination: he may be able to
say
what the experiment died of. ~ Sir Ronald Aylmer Fisher
The plural of anecdote is not data. ~ Roger Brinner
The combination of some data and an aching desire for an answer does not
ensure that a reasonable answer can be extracted from a given body of
data.
~ John Tukey

2015-08-07 21:27 GMT+02:00 Joaqu?n Aldabe <joaquin.aldabe at gmail.com>:

Hi all, I?m running a mixed model with poisson error distribution as I

have

count data of a shorebird species. My predictors are AMGP (another
shorebird that is usually associated to "my" species and probably has
ecological influence on it), grass height and field distance to the

lagoon.

I?m having a hard time trying to interpret second order interactions. I
appreciate your comments of references recommendations. Here is the

output

of my model. All the best, Joaqu?n.

Call:glmmadmb(formula = BBSA ~ AMGP * Grass_height *
Distance_to_lagoon +     (1 | Field_name) + offset(log(Field_area.o)),
data = my4S,     family = "nbinom1", zeroInflation = TRUE)
AIC: 226
Coefficients:                                     Estimate Std. Error
z value Pr(>|z|)    (Intercept)                           -5.8797
1.2250   -4.80  1.6e-06 ***AMGP
3.3086     1.0246    3.23   0.0012 ** Grass_height
     -4.1592     1.9961   -2.08   0.0372 *  Distance_to_lagoon
           -0.5029     1.3929   -0.36   0.7180    AMGP:Grass_height
                  2.9442     1.5142    1.94   0.0518 .
AMGP:Distance_to_lagoon               -2.4695     0.9632   -2.56
0.0104 *  Grass_height:Distance_to_lagoon        0.0339     2.0065
0.02   0.9865    AMGP:Grass_height:Distance_to_lagoon  -4.0363
1.3264   -3.04   0.0023 ** ---Signif. codes:  0 ?***? 0.001 ?**? 0.01
?*? 0.05 ?.? 0.1 ? ? 1
Number of observations: total=64, Field_name=16 Random effect
variance(s):Group=Field_name            Variance StdDev(Intercept)
0.03721 0.1929
Negative binomial dispersion parameter: 20.633 (std. err.:
7.4042)Zero-inflation: 1e-06  (std. err.:  6.8336e-09 )
Log-likelihood: -102


--
*Joaqu?n Aldabe*

*Grupo Biodiversidad, Ambiente y Sociedad*
Centro Universitario de la Regi?n Este, Universidad de la Rep?blica
Ruta 15 (y Ruta 9), Km 28.500, Departamento de Rocha

*Departamento de Conservaci?n*
Aves Uruguay
BirdLife International
Canelones 1164, Montevideo

https://sites.google.com/site/joaquin.aldabe
<https://sites.google.com/site/perfilprofesionaljoaquinaldabe>

        [[alternative HTML version deleted]]

        [[alternative HTML version deleted]]

--

*Ken Beath*
Lecturer
Statistics Department
MACQUARIE UNIVERSITY NSW 2109, Australia

Phone: +61 (0)2 9850 8516

Level 2, AHH
http://stat.mq.edu.au/our_staff/staff_-_alphabetical/staff/beath,_ken/

CRICOS Provider No 00002J
This message is intended for the addressee named and m...{{dropped:12}}

Consider taking a look at
http://ms.mcmaster.ca/~bolker/R/misc/foxchapter/bolker_chap.html .

On Mon, Aug 10, 2015 at 10:30 AM, Joaqu?n Aldabe
<joaquin.aldabe at gmail.com> wrote:

Thankyoy very much Ken and Thierry. I'll try that. Is there any specific
function for making graphs of the model? I appreciate recommendations or
citations to explore how to graph model results. Thanks again. Joaquin
El 10/08/2015 07:36, "Ken Beath" <ken.beath at mq.edu.au> escribi?:

There is actually a continuous third order interaction, which are just
about impossible to interpret.

As well as the graphs, which you always should start with, think about
what is sensible for the model. Should any of the predictors be

log-scaled?

As AMGP is a count it wouldn't be surprising it it is, or use log(1+x)

if

there are zeroes.

Look at diagnostic plots and make sure that everything is fine,

especially

that there are no outliers which might be causing the interaction.

Ken

On 10 August 2015 at 17:02, Thierry Onkelinx <thierry.onkelinx at inbo.be>
wrote:

Dear Joaquin,

Your mail is mangled because you send it in HTML. This makes the model
output very hard to read.

It's impossible to interpret model output without thorough knowledge of
the
dataset. I suggest that you try to make some graphs of the model.

Best regards,

ir. Thierry Onkelinx
Instituut voor natuur- en bosonderzoek / Research Institute for Nature

and

Forest
team Biometrie & Kwaliteitszorg / team Biometrics & Quality Assurance
Kliniekstraat 25
1070 Anderlecht
Belgium

To call in the statistician after the experiment is done may be no more
than asking him to perform a post-mortem examination: he may be able to
say
what the experiment died of. ~ Sir Ronald Aylmer Fisher
The plural of anecdote is not data. ~ Roger Brinner
The combination of some data and an aching desire for an answer does

not

ensure that a reasonable answer can be extracted from a given body of
data.
~ John Tukey

2015-08-07 21:27 GMT+02:00 Joaqu?n Aldabe <joaquin.aldabe at gmail.com>:

Hi all, I?m running a mixed model with poisson error distribution as

I

have

count data of a shorebird species. My predictors are AMGP (another
shorebird that is usually associated to "my" species and probably has
ecological influence on it), grass height and field distance to the

lagoon.

I?m having a hard time trying to interpret second order

interactions. I

appreciate your comments of references recommendations. Here is the

output

of my model. All the best, Joaqu?n.

Call:glmmadmb(formula = BBSA ~ AMGP * Grass_height *
Distance_to_lagoon +     (1 | Field_name) +

offset(log(Field_area.o)),

data = my4S,     family = "nbinom1", zeroInflation = TRUE)
AIC: 226
Coefficients:                                     Estimate Std. Error
z value Pr(>|z|)    (Intercept)                           -5.8797
1.2250   -4.80  1.6e-06 ***AMGP
3.3086     1.0246    3.23   0.0012 ** Grass_height
     -4.1592     1.9961   -2.08   0.0372 *  Distance_to_lagoon
           -0.5029     1.3929   -0.36   0.7180    AMGP:Grass_height
                  2.9442     1.5142    1.94   0.0518 .
AMGP:Distance_to_lagoon               -2.4695     0.9632   -2.56
0.0104 *  Grass_height:Distance_to_lagoon        0.0339     2.0065
0.02   0.9865    AMGP:Grass_height:Distance_to_lagoon  -4.0363
1.3264   -3.04   0.0023 ** ---Signif. codes:  0 ?***? 0.001 ?**? 0.01
?*? 0.05 ?.? 0.1 ? ? 1
Number of observations: total=64, Field_name=16 Random effect
variance(s):Group=Field_name            Variance StdDev(Intercept)
0.03721 0.1929
Negative binomial dispersion parameter: 20.633 (std. err.:
7.4042)Zero-inflation: 1e-06  (std. err.:  6.8336e-09 )
Log-likelihood: -102


--
*Joaqu?n Aldabe*

*Grupo Biodiversidad, Ambiente y Sociedad*
Centro Universitario de la Regi?n Este, Universidad de la Rep?blica
Ruta 15 (y Ruta 9), Km 28.500, Departamento de Rocha

*Departamento de Conservaci?n*
Aves Uruguay
BirdLife International
Canelones 1164, Montevideo

https://sites.google.com/site/joaquin.aldabe
<https://sites.google.com/site/perfilprofesionaljoaquinaldabe>

        [[alternative HTML version deleted]]

        [[alternative HTML version deleted]]

--

*Ken Beath*
Lecturer
Statistics Department
MACQUARIE UNIVERSITY NSW 2109, Australia

Phone: +61 (0)2 9850 8516

Level 2, AHH
http://stat.mq.edu.au/our_staff/staff_-_alphabetical/staff/beath,_ken/

CRICOS Provider No 00002J
This message is intended for the addressee named and m...{{dropped:12}}

Thankyou Ben! It looks great. Cheers
El 10/08/2015 11:54, "Ben Bolker" <bbolker at gmail.com> escribi?:

Consider taking a look at
http://ms.mcmaster.ca/~bolker/R/misc/foxchapter/bolker_chap.html .

On Mon, Aug 10, 2015 at 10:30 AM, Joaqu?n Aldabe
<joaquin.aldabe at gmail.com> wrote:

Thankyoy very much Ken and Thierry. I'll try that. Is there any specific
function for making graphs of the model? I appreciate recommendations or
citations to explore how to graph model results. Thanks again. Joaquin
El 10/08/2015 07:36, "Ken Beath" <ken.beath at mq.edu.au> escribi?:

There is actually a continuous third order interaction, which are just
about impossible to interpret.

As well as the graphs, which you always should start with, think about
what is sensible for the model. Should any of the predictors be

log-scaled?

As AMGP is a count it wouldn't be surprising it it is, or use log(1+x)

if

there are zeroes.

Look at diagnostic plots and make sure that everything is fine,

especially

that there are no outliers which might be causing the interaction.

Ken

On 10 August 2015 at 17:02, Thierry Onkelinx <thierry.onkelinx at inbo.be

wrote:

Dear Joaquin,

Your mail is mangled because you send it in HTML. This makes the model
output very hard to read.

It's impossible to interpret model output without thorough knowledge

of

the
dataset. I suggest that you try to make some graphs of the model.

Best regards,

ir. Thierry Onkelinx
Instituut voor natuur- en bosonderzoek / Research Institute for

Nature and

Forest
team Biometrie & Kwaliteitszorg / team Biometrics & Quality Assurance
Kliniekstraat 25
1070 Anderlecht
Belgium

To call in the statistician after the experiment is done may be no

more

than asking him to perform a post-mortem examination: he may be able

to

say
what the experiment died of. ~ Sir Ronald Aylmer Fisher
The plural of anecdote is not data. ~ Roger Brinner
The combination of some data and an aching desire for an answer does

not

ensure that a reasonable answer can be extracted from a given body of
data.
~ John Tukey

2015-08-07 21:27 GMT+02:00 Joaqu?n Aldabe <joaquin.aldabe at gmail.com>:

Hi all, I?m running a mixed model with poisson error distribution

as I

have

count data of a shorebird species. My predictors are AMGP (another
shorebird that is usually associated to "my" species and probably

has

ecological influence on it), grass height and field distance to the

lagoon.

I?m having a hard time trying to interpret second order

interactions. I

appreciate your comments of references recommendations. Here is the

output

of my model. All the best, Joaqu?n.

Call:glmmadmb(formula = BBSA ~ AMGP * Grass_height *
Distance_to_lagoon +     (1 | Field_name) +

offset(log(Field_area.o)),

data = my4S,     family = "nbinom1", zeroInflation = TRUE)
AIC: 226
Coefficients:                                     Estimate Std.

Error

z value Pr(>|z|)    (Intercept)                           -5.8797
1.2250   -4.80  1.6e-06 ***AMGP
3.3086     1.0246    3.23   0.0012 ** Grass_height
     -4.1592     1.9961   -2.08   0.0372 *  Distance_to_lagoon
           -0.5029     1.3929   -0.36   0.7180    AMGP:Grass_height
                  2.9442     1.5142    1.94   0.0518 .
AMGP:Distance_to_lagoon               -2.4695     0.9632   -2.56
0.0104 *  Grass_height:Distance_to_lagoon        0.0339     2.0065
0.02   0.9865    AMGP:Grass_height:Distance_to_lagoon  -4.0363
1.3264   -3.04   0.0023 ** ---Signif. codes:  0 ?***? 0.001 ?**?

0.01

?*? 0.05 ?.? 0.1 ? ? 1
Number of observations: total=64, Field_name=16 Random effect
variance(s):Group=Field_name            Variance StdDev(Intercept)
0.03721 0.1929
Negative binomial dispersion parameter: 20.633 (std. err.:
7.4042)Zero-inflation: 1e-06  (std. err.:  6.8336e-09 )
Log-likelihood: -102


--
*Joaqu?n Aldabe*

*Grupo Biodiversidad, Ambiente y Sociedad*
Centro Universitario de la Regi?n Este, Universidad de la Rep?blica
Ruta 15 (y Ruta 9), Km 28.500, Departamento de Rocha

*Departamento de Conservaci?n*
Aves Uruguay
BirdLife International
Canelones 1164, Montevideo

https://sites.google.com/site/joaquin.aldabe
<https://sites.google.com/site/perfilprofesionaljoaquinaldabe>

        [[alternative HTML version deleted]]

        [[alternative HTML version deleted]]

--

*Ken Beath*
Lecturer
Statistics Department
MACQUARIE UNIVERSITY NSW 2109, Australia

Phone: +61 (0)2 9850 8516

Level 2, AHH
http://stat.mq.edu.au/our_staff/staff_-_alphabetical/staff/beath,_ken/

CRICOS Provider No 00002J
This message is intended for the addressee named and m...{{dropped:12}}