Skip to content
Prev 4197 / 20628 Next

Multi-level qualitative (fixed-effects) factors

Sorry to everyone,

I guess i should rephrase this completely as i think i am not getting across what i want to ( my fault, i apologise for the spam!).

I am trying to discover which traits correlate with threatened or not (binary response- i do have different levele of threat but am unsure how to model continuos response variables in GLMM - quasi poisson?)

I shall work through a quick example and would appreciate it if you could tell me if my conclusions are justified.

Lets just say for ease i am have two models -  i can send str(traits) etc if required -
This shows that with habit the model has a AIC value of 1406 and habit2 has the significant effect on the threat status of a species?

Now model 2
This shows that adding breeding system to the model doesn't fit the data any better (the AIC is higher), therefore breedingsystem has no significant effect on the threat status of a species

Under this simple example i can conclude that the habit of a species is significant in determining how threatened the species is and in particular species from habit2 are more threatened than those not from habit2.

Are these inferences justified or am i missing something?

Thanks
On 3 Aug 2010, at 08:50, Andrew Dolman wrote:
I don't get it. How can you fit the model with just 1 of three levels
of factor "habitat" and have the same number of observations as when
you run the model with all three? (It must have at least 2 levels to
fit anyway) Also, in the first example you have 4 levels of habitat.

Are they different levels of habitat resolution? e.g.

Aquatic - non aquatic
Aquatic - Terrestrial - Epiphytic
Aquatic - Terrestrial - Epiphytic - Up elephant's noses


Please read the posting guide and include proper examples of what you
are doing and what the data look like.


andydolman at gmail.com
On 3 August 2010 08:48, Peter Francis <peterfrancis at me.com> wrote: