[R] Mixed models fixed effects
ONKELINX, Thierry
Thierry.ONKELINX at inbo.be
Fri Mar 13 11:21:20 CET 2009
Dear Emma,
First of all make shure that style and habitat are factors (assuming
that they are categorical).
glmer(passes~hvolume+style+habitat(1|Site),family = poisson)
Style has 3 levels, habitat 5 levels. So your model needs to estimate 1
parameter for hvolume, 2 for style, 4 for habitat and 1 for site. In
total 8 parameters. A rule of thumbs tells you that you need about 10
samples for each parameter. Hence you dataset (n = 48) is to small for
this kind of model.
Aditionally the variable habitat has some rare levels. Only one occures
of levels 3 and 5, and just a few more for levels 2 and 4. This kind of
data won't give you reliable estimates for habitat. So I would suggest
to drop this from your model.
HTH,
Thierry
------------------------------------------------------------------------
----
ir. Thierry Onkelinx
Instituut voor natuur- en bosonderzoek / Research Institute for Nature
and Forest
Cel biometrie, methodologie en kwaliteitszorg / Section biometrics,
methodology and quality assurance
Gaverstraat 4
9500 Geraardsbergen
Belgium
tel. + 32 54/436 185
Thierry.Onkelinx at inbo.be
www.inbo.be
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
-----Oorspronkelijk bericht-----
Van: Emma Stone [mailto:Emma.Stone at bristol.ac.uk]
Verzonden: vrijdag 13 maart 2009 11:08
Aan: ONKELINX, Thierry; Emma Stone; r-help at r-project.org
Onderwerp: RE: [R] Mixed models fixed effects
Hi Thierry,
Thanks, my data are pasted in below: n = 48, with different numbers if
site
replicates.
Site passes length width height ohwidth ohheight style shape nogap
pgaps
1 1 32 38.21 7.18 2.60 0.00 0.00 1 1 0
0.00
2 1 7 154.38 2.55 2.43 0.00 0.00 2 1 0
0.00
3 2 24 130.00 2.73 4.96 5.83 2.27 1 2 0
1.92
4 2 43 130.00 2.73 4.96 3.45 2.45 1 2 0
1.91
5 2 32 130.00 3.30 4.34 3.67 2.63 3 2 0
0.00
6 2 9 130.00 3.30 4.34 2.13 1.53 2 2 0
0.00
7 2 25 214.00 1.65 6.23 2.80 2.53 3 2 0
0.00
8 2 1 214.00 2.20 1.05 0.00 0.00 2 1 0
0.00
9 2 57 188.00 5.87 4.97 3.23 2.13 2 2 0
0.00
10 2 32 211.00 2.63 6.09 4.37 2.50 2 2 0
0.00
11 2 14 211.00 2.63 6.09 3.50 2.53 2 2 0
0.00
12 3 78 192.00 2.30 2.40 0.33 0.88 2 2 0
2.19
13 3 11 192.00 2.50 2.50 0.13 0.55 2 2 0
2.19
14 3 9 444.00 2.10 1.93 0.00 0.00 2 1 2
6.19
15 4 40 44.00 1.83 2.05 0.00 0.00 1 1 1
31.81
16 4 112 44.00 1.58 1.73 0.00 0.00 1 1 1
31.81
17 4 23 98.00 2.18 1.70 0.00 0.00 1 3 0
2.55
18 4 23 98.00 1.78 1.60 0.00 0.00 1 1 0
2.55
19 4 4 84.00 1.73 1.98 0.00 0.00 1 3 1
17.02
20 4 72 88.00 3.30 5.30 3.30 1.45 1 2 0
0.00
21 5 207 116.00 3.08 3.38 3.65 3.68 1 2 0
0.00
22 5 2 116.00 2.85 2.18 3.18 3.53 1 2 0
0.00
23 5 81 104.00 2.68 2.78 1.00 0.75 2 1 0
0.00
24 5 3 104.00 2.68 2.38 0.00 0.00 2 1 0
0.00
25 5 21 59.00 2.78 4.03 2.23 1.75 2 2 0
6.78
26 5 1 59.00 2.78 3.33 1.38 1.35 2 2 0
6.78
27 6 82 165.56 2.50 4.00 1.00 1.33 1 2 0
0.00
28 6 1 165.56 2.60 3.63 0.50 1.40 1 2 0
0.00
29 6 46 69.35 3.23 10.67 4.73 4.67 1 2 0
0.00
30 6 7 165.56 3.50 10.17 2.67 1.67 1 2 0
0.00
31 6 45 136.00 3.33 3.67 0.00 0.00 1 2 0
2.20
32 6 3 136.00 3.33 2.75 0.00 0.00 1 2 0
3.30
33 6 34 82.00 2.83 12.00 3.33 3.83 1 2 0
0.00
34 7 0 228.00 2.57 2.70 0.30 1.00 2 2 0
0.50
35 7 148 228.00 3.08 4.68 0.75 1.75 2 2 0
0.05
36 7 0 208.00 2.40 2.90 0.00 0.00 2 1 0
1.44
37 7 38 208.00 2.57 2.63 0.00 0.00 2 1 0
1.44
38 7 0 112.00 2.13 1.65 0.00 0.00 2 1 0
0.50
39 7 0 192.00 1.55 1.55 0.00 0.00 2 1 0
0.00
40 7 0 132.00 1.56 1.41 0.00 0.00 2 1 0
0.00
41 8 52 148.00 5.05 7.62 2.55 3.38 2 2 0
0.00
42 8 18 148.00 4.08 5.80 0.65 0.63 2 2 0
0.00
43 8 19 210.00 3.08 4.58 0.70 0.88 2 2 0
1.42
44 8 0 210.00 3.68 4.90 0.25 0.38 3 2 0
1.42
45 8 3 246.00 2.18 2.43 0.00 0.00 2 1 3
15.24
46 8 0 246.00 1.93 2.20 0.00 0.00 2 1 3
15.24
47 8 3 142.00 1.60 2.03 0.00 0.00 1 1 3
26.19
48 8 0 142.00 1.37 1.87 0.00 0.00 1 1 3
26.19
habitat side oharea hvolume
1 1 2 0.00 713.30
2 1 4 0.00 956.62
3 1 2 1720.43 1760.30
4 1 4 1098.83 1760.30
5 1 4 1254.77 1861.86
6 1 2 423.66 1861.86
7 1 1 1515.98 2199.81
8 1 3 0.00 494.34
9 1 2 1293.42 5484.69
10 1 3 2305.18 3379.52
11 1 1 1868.41 3379.52
12 1 2 55.76 1059.84
13 1 4 13.73 1200.00
14 1 1 0.00 1799.53
15 1 1 0.00 165.07
16 1 3 0.00 120.27
17 1 2 0.00 363.19
18 1 4 0.00 279.10
19 1 4 0.00 287.73
20 1 2 421.08 1539.12
21 1 2 1558.11 1207.61
22 2 4 1302.15 720.71
23 2 2 78.00 774.84
24 1 4 0.00 663.35
25 1 2 230.25 661.00
26 1 4 109.92 546.19
27 1 3 220.19 1655.60
28 3 1 115.89 1562.56
29 1 1 1531.88 2390.09
30 1 3 738.22 5893.11
31 2 2 0.00 1662.07
32 4 4 0.00 1245.42
33 5 3 1045.82 2784.72
34 1 2 68.40 1582.09
35 1 4 299.25 3286.48
36 1 2 0.00 1447.68
37 1 4 0.00 1405.89
38 4 3 0.00 393.62
39 4 1 0.00 461.28
40 4 1 0.00 290.35
41 1 4 1275.61 5695.19
42 1 2 60.61 3502.27
43 1 1 129.36 2962.34
44 1 3 19.95 3786.72
45 1 1 0.00 1303.16
46 1 3 0.00 1044.52
47 1 2 0.00 461.22
48 1 4 0.00 363.79
--On 13 March 2009 11:01 +0100 "ONKELINX, Thierry"
<Thierry.ONKELINX at inbo.be> wrote:
> Dear Emma,
>
> Have you tried a simpler model? False convergence can be due to an
> overcomplex model. Can you give a brief outline of your data? E.g. how
> many sites, how many data per site, ... Cross tabulations of all pairs
> of factor variables are usefull too.
>
> HTH,
>
> Thierry
>
>
>
------------------------------------------------------------------------
> ----
> ir. Thierry Onkelinx
> Instituut voor natuur- en bosonderzoek / Research Institute for Nature
> and Forest
> Cel biometrie, methodologie en kwaliteitszorg / Section biometrics,
> methodology and quality assurance
> Gaverstraat 4
> 9500 Geraardsbergen
> Belgium
> tel. + 32 54/436 185
> Thierry.Onkelinx at inbo.be
> www.inbo.be
>
> 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
>
> -----Oorspronkelijk bericht-----
> Van: Emma Stone [mailto:Emma.Stone at bristol.ac.uk]
> Verzonden: vrijdag 13 maart 2009 10:50
> Aan: ONKELINX, Thierry; Emma Stone; r-help at r-project.org
> Onderwerp: RE: [R] Mixed models fixed effects
>
> Hi Thierry,
>
> That's great thanks!
>
> I have done as you have said but I keep getting a warning message here
> is
> my code:
>
> G1Hvol<-glmer(passes~hvolume+style+habitat(1|Site),family = poisson)
>
> And this is the message i get:
> Warning message:
> In mer_finalize(ans) : false convergence (8)
>
> any ideas??
>
> Emma
>
>
> --On 11 March 2009 15:45 +0100 "ONKELINX, Thierry"
> <Thierry.ONKELINX at inbo.be> wrote:
>
>> Hi Emma,
>>
>> Continuous predictors are no problem at all. You can mix both
> continuous
>> and categorial predictors if needed. I suppose your response are
> counts
>> (the number of bats that passes)? In that case a generalised linear
>> mixed model is more appropriate. With the lme4 package you could try
>> something like this:
>>
>> library(lme4)
>> Model <- glmer(BatPasses ~ Width + Height + (1|Site), family =
> poisson)
>>
>> HTH,
>>
>> Thierry
>>
>> PS There is a mailing list dedicated to mixed models:
> R-Sig-MixedModels
>>
>
------------------------------------------------------------------------
>> ----
>> ir. Thierry Onkelinx
>> Instituut voor natuur- en bosonderzoek / Research Institute for
Nature
>> and Forest
>> Cel biometrie, methodologie en kwaliteitszorg / Section biometrics,
>> methodology and quality assurance
>> Gaverstraat 4
>> 9500 Geraardsbergen
>> Belgium
>> tel. + 32 54/436 185
>> Thierry.Onkelinx at inbo.be
>> www.inbo.be
>>
>> 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
>>
>> -----Oorspronkelijk bericht-----
>> Van: r-help-bounces at r-project.org
> [mailto:r-help-bounces at r-project.org]
>> Namens Emma Stone
>> Verzonden: woensdag 11 maart 2009 15:29
>> Aan: r-help at r-project.org
>> Onderwerp: Re: [R] Mixed models fixed effects
>>
>> Dear All,
>>
>> This may sound like a dumb question but I am trying to use a mixed
> model
>> to
>> determine the predictors of bat activity along hedges within 8 sites.
> So
>> my
>> response is continuous (bat passes) my predictors fixed effects are
>> continuous (height metres), width (metres) etc and the random effect
> is
>> site - can you tell me if the fixed effects can be continuous as all
>> the
>> examples I have read show them as categorical, but this is not
covered
>> in
>> any documents I can find.
>>
>> Help!
>>
>> Emma
>>
>> ______________________________________________
>> 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.
>>
>> Dit bericht en eventuele bijlagen geven enkel de visie van de
> schrijver
>> weer en binden het INBO onder geen enkel beding, zolang dit bericht
> niet
>> bevestigd is door een geldig ondertekend document. The views
expressed
> in
>> this message and any annex are purely those of the writer and may
not
> be
>> regarded as stating an official position of INBO, as long as the
> message
>> is not confirmed by a duly signed document.
>
>
>
> ----------------------
> Emma Stone
> Postgraduate Researcher
> Bat Ecology and Bioacoustics Lab
> & Mammal Research Unit
> School of Biological Sciences,
> University of Bristol, Woodland Road,
> Bristol, BS8 1UG
> Email: emma.stone at bristol.ac.uk
>
>
>
> Dit bericht en eventuele bijlagen geven enkel de visie van de
schrijver
> weer en binden het INBO onder geen enkel beding, zolang dit bericht
niet
> bevestigd is door een geldig ondertekend document. The views expressed
in
> this message and any annex are purely those of the writer and may not
be
> regarded as stating an official position of INBO, as long as the
message
> is not confirmed by a duly signed document.
----------------------
Emma Stone
Postgraduate Researcher
Bat Ecology and Bioacoustics Lab
& Mammal Research Unit
School of Biological Sciences,
University of Bristol, Woodland Road,
Bristol, BS8 1UG
Email: emma.stone at bristol.ac.uk
Dit bericht en eventuele bijlagen geven enkel de visie van de schrijver weer
en binden het INBO onder geen enkel beding, zolang dit bericht niet bevestigd is
door een geldig ondertekend document. The views expressed in this message
and any annex are purely those of the writer and may not be regarded as stating
an official position of INBO, as long as the message is not confirmed by a duly
signed document.
More information about the R-help
mailing list