LMM: including ranef or not?

An embedded and charset-unspecified text was scrubbed...
Name: not available
URL: <https://stat.ethz.ch/pipermail/r-sig-mixed-models/attachments/20140721/890512b0/attachment.pl>
If you're going to use ud95 as the response you might as well average
the prop_land values per bird (i.e., aggregate the data down to a single
data record per bird); the within-bird variation in prop_land values
won't affect the model output at all (although the _number_ of
observations per bird will; if you have unbalanced information in this
way, you should incorporate weights proportional to the number of
observations as well).

  With only 6 colonies you're going to have some difficulty estimating
the among-colony variance very well; if you end up with zero estimates
of among-colony variance, you might want to use blme or set the colonies
as fixed effects ...

  If you use prop_land as the response variable you would indeed want to
put bird_id in as a random effect (and you might consider estimating the
proportional as a binomial (GLMM) response, _if_ you know the total
number of fixes for each bird)
Dear Christian,

thanks for your answer. So, including the ranef or not depends on the 
response, doesn't it? If I tested the model the other way around 
(prop_land~ud95...) I would have to include bird_id as ranef, because 
there are more than one measurments for each bird, right?

Best,
Anna

Am 21.07.2014 13:49, schrieb Christian Ritz:
Dear Anna,

no, with only one measurement per bird there is no need for a 
bird-specific random effect.

Your model looks okay to me.

Best wishes Christian

On 21-07-2014 13:37, Anna-Marie Corman wrote:
Dear list,

I want to test whether the UD sizes of several tracked seabird
individuals from 6 different breeding  colonies depends on the foraging
target (on land or at sea) as  follows:

mod<-lmer(ud95~prop_land+(1|colony),data=dat);

I have one UD95 value for each individual, but the proportion of fixes
on land are on a trip basis, i.e. several values for one individual. Do
I need to include bird_id as random factor to exclude pseudo
replication, though the response has only one value per individual???

Many thanks in advance.

Best,
Anna

data:

     bird_id colony FT_id year sex max_speed max_distnest  tot_dist       tdur mean_dist mean_distnest
1 HA1_2012  Amrum     1 2012   1  53.65358    36.008004 101.34099  7.5827778 0.4444780     15.867350
2 HA1_2012  Amrum     2 2012   1  63.88851    69.993403 149.00254  6.1788889 0.8186953     46.947190
3 HA1_2012  Amrum     3 2012   1  82.68318    70.532407 176.65181  7.7008333 0.7886241     48.160436
4 HA1_2012  Amrum     4 2012   1  56.25961     5.293994  15.11632  0.8130556 0.6298466      3.710259
5 HA1_2012  Amrum     5 2012   1  70.04150    71.002162 215.32017 12.6369444 0.5851092     42.360905
6 HA1_2012  Amrum     6 2012   1  71.40167    71.712123 213.43533 12.1355556 0.5995374     54.878232
    mean_speed  prop_sea   prop_land straightness   prop_day prop_night ft_start ft_start_s   ft_end ft_end_s
1   13.01700 0.9912664 0.008733624    0.3553153 0.08296943  0.9170306 21:34:09      77649 05:11:03    18663
2   25.24134 0.2786885 0.721311500    0.4697464 1.00000000  0.0000000 06:47:53      24473 13:00:35    46835
3   26.50082 0.1644444 0.835555600    0.3992736 0.92888890  0.0711111 04:28:00      16080 12:12:06    43926
4   22.27650 0.8000000 0.200000000    0.3502171 0.24000000  0.7600000 04:21:24      15684 05:12:12    18732
5   19.70902 0.3821138 0.617886200    0.3297516 0.84552850  0.1544715 03:05:05      11105 15:45:16    56716
6   22.07548 0.2997199 0.700280100    0.3359899 0.91596640  0.0840336 04:02:14      14534 16:12:19    58339

      ud95 ud50 id30  rd30 udoi50_colmean udoi95_colmean idoi30_colmean rdoi30_colmean
1 460.73 9.76 24.8 64.16          0.002         0.0985          1e-04         0.0059
2 460.73 9.76 24.8 64.16          0.002         0.0985          1e-04         0.0059
3 460.73 9.76 24.8 64.16          0.002         0.0985          1e-04         0.0059
4 460.73 9.76 24.8 64.16          0.002         0.0985          1e-04         0.0059
5 460.73 9.76 24.8 64.16          0.002         0.0985          1e-04         0.0059
6 460.73 9.76 24.8 64.16          0.002         0.0985          1e-04         0.0059

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

	[[alternative HTML version deleted]]

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

Dear Ben and Christian,

thanks again for your help! I know the number of fixes per bird, though 
my dataset is on a trip basis. All individuals made a different number 
of trips during the study period.

I will try this out.

Best,
Anna

Am 21.07.2014 23:24, schrieb Ben Bolker:
   If you're going to use ud95 as the response you might as well average
the prop_land values per bird (i.e., aggregate the data down to a single
data record per bird); the within-bird variation in prop_land values
won't affect the model output at all (although the _number_ of
observations per bird will; if you have unbalanced information in this
way, you should incorporate weights proportional to the number of
observations as well).

   With only 6 colonies you're going to have some difficulty estimating
the among-colony variance very well; if you end up with zero estimates
of among-colony variance, you might want to use blme or set the colonies
as fixed effects ...

   If you use prop_land as the response variable you would indeed want to
put bird_id in as a random effect (and you might consider estimating the
proportional as a binomial (GLMM) response, _if_ you know the total
number of fixes for each bird)