[R] function effect and standard error

[John Fox]

Dear Emilie,

This is, I think, the effect() function in the effects() package.

By the way, the model that you've fit with glm() is just a linear
model, and could also have been fit with lm().

For a model with this simple structure, effect() computes the "adjusted
means" for the factor sp, holding other predictors to their average
values. These effects are just fitted values under the model, and the
standard errors reported are for the fitted values.

For details, see the paper at
<http://www.jstatsoft.org/counter.php?id=75&url=v08/i15/effect-displays-revised.pdf>.

I hope this helps.

John

On Tue, 1 Nov 2005 12:55:17 -0500
 "Emilie Berthiaume" <emilie.berthiaume at usherbrooke.ca> wrote:
> Hi list!
> 
> I did the following regression:
> reg1 <-  glm(alti~sp + ovent + vivent + nuage, family=gaussian,
> data=meteo1)
> 
> I was interested in knowing the effect of the species (sp) in reg1
> and so I used the function «effect»:
> 
> effect.sp <- effect ("sp", reg1, se=TRUE)
> 
> with this output:
> sp
>       AK       BW       NH       OS       RT       SS 
> 2.730101 2.885363 2.753774 2.750311 3.084606 2.834390 
> 
> If I enter the following command:
>     effect.sp$se
> I get this output:
>             1        253       3100        488        514       6100 
>     0.07924610 0.06713200 0.11493178 0.13106639 0.05252749 0.04208334
> 
> 
> My question is:  Do the numbers on the second line of this output
> represent the standard error?  What do the numbers on the top line
> represent?
> 
> Thank you,
> 
> Emilie Berthiaume
> graduate student
> Biology Departement
> Université de Sherbrooke
> 2500 boul. de l'Université
> Sherbrooke, Québec
> J1K 2R1 CANADA
> 
> Tél: 1-819-821-8000 poste 2059
> Fax: 1-819-821-8049
> emilie.berthiaume at USherbrooke.ca
> 	[[alternative HTML version deleted]]
> 

--------------------------------
John Fox
Department of Sociology
McMaster University
Hamilton, Ontario, Canada
http://socserv.mcmaster.ca/jfox/