[R-sig-ME] binary response, random factor
Boris Fuchs
boris.fuchs at schultraining.ch
Fri Mar 6 10:07:45 CET 2015
-----BEGIN PGP SIGNED MESSAGE-----
Hash: SHA1
Can you please send this to r-sig-mixed-models at r-project.org ?
I think the answer would be of general interest.
I do have one paper (Pasch et al Am Nat 2013 in which we initially
considered using a mixed model but decided not to in the end -- you could
cite that in support ...
On 15-03-05 11:05 AM, Boris Fuchs wrote:
> Hi Ben
>
>
>
> Thirst: thanks for all the answers you share in the net.
>
>
>
> Guess you are busy. I try to be short.
>
>
>
> I try to publish my master thesis:
>
>
>
> I have a data set of 106 Scandinavian wolves. A binary response
> whether they are tested positive or negative for a parasite (Sarcoptic
> mange). I test if individual variables (sex, age
) or environmental
> variables (fox density, latitude
) are better explaining the test
> results (positive or negative).
>
>
>
> I use glms, selected by AICc in R.
>
>
>
>
>
> 16 are positive.
>
>
>
> I am aware of potential overfitting problems (max K = n/10 or better
> 20
)
>
>
>
> About 40 individuals got sampled >1 up to 4 times. > random factor
>
>
>
> Some territories are sampled multiple times. à random factor
>
>
>
> However:
>
> I did not succeed with building random structures and I belief because
> of missing variability within the random factors. Most of the
> individuals are tested only once. Easier if most of them would be
> tested several times and would be both, positive and negative.
> The same with the territories. Is this right?
>
>
>
> I would like to state the lacking variation as a reason to not use a
> random structure. But I need some published support (not to say it
> works anyway, but not). I seem not skilled enough to understand
> (select) the key literature. Do you have any hint?
>
>
>
>
>
> Since most of the individuals are sampled once and the time between
> the sampling is by far enough to change from positive to negative and
> vica versa I think I can take the risk of violation of independence.
>
>
>
>
>
>
>
> Boris
>
>
>
>
>
>
>
> boris fuchs ole evenstadsvei 1156 2480 koppang +47 41 51 49 30
> boris.fuchs at schultraining.ch
>
>
>
>
-----BEGIN PGP SIGNATURE-----
Version: GnuPG v1.4.11 (GNU/Linux)
iQEcBAEBAgAGBQJU+L7dAAoJEOCV5YRblxUH1IUH/jciVrXp/i5o6B8w6rM68LEy
CMHD3udnNLr+DFBXKVjH+XM6pljGDQSBUKBYncSZgY0jTWDsMfHTDgt0mHttlodF
koAebeqminRsvp3yc+lgpE4stBAEVUPRIh2onaVS6pzISYaoZEmmGwrJNAerZvyl
sEUgTH+pLSsSdXDJzKR8KYgocoAifMvHoDmwjtkdIz6FTdGxpqgkHNjLHj5Itesd
mexuWPyFFcmHI1IuCUNiuZnJNvA07c1isWSncuY0+NkfsRZWjjTNi8PNSuClLDXx
9K0k+8HySbH9zNqyFod9ErDd566AVeSVgqItA0xEtNVE6puyvFGF1tT1gw5WPes=
=15/3
-----END PGP SIGNATURE-----
boris fuchs
ole evenstadsvei 1156
2480 koppang
+47 41 51 49 30
boris.fuchs at schultraining.ch
[[alternative HTML version deleted]]
More information about the R-sig-mixed-models
mailing list