In lm depend to research question we can get output by summary(model) and summary.aov(model). In the first condition we can get details for lm with estimate and t value (the subgroups indicated in Bold):
For instance interaction of moisture (dry, moist,wet) and species(acetosa,acetosella, maritimus) would be like following:
> summary(model1)
Coefficients:
Estimate Std. Error t value Pr(>|t|)
(Intercept) 27.46217 0.99102 27.711 < 2e-16 ***
speciesacetosa -24.74597 1.03466 -23.917 < 2e-16 ***
speciesmaritimus -3.77972 1.03879 -3.639 0.000299 ***
moisturemoist -1.32010 1.04714 -1.261 0.207936
moisturewet -2.99489 1.03908 -2.882 0.004096 **
......
we can also get ANOVA output with summary.aov(model) indicating main effects and their interactions without subgroups:
>summary.aov(model1)
Df Sum Sq Mean Sq F value Pr(>F)
species 2 49261 24630 2410.690 < 2e-16 ***
moisture 2 463 231 22.652 3.39e-10 ***
soil 2 182 91 8.883 0.000159 ***
......
For lm it is clear what to do but how we can get outputs without details and levels in the GLM and GAM ? I mean if i want to know about only main effect of moisture, species and their interactions which simple method do you recommend?
Warm regards,
Mehdi
For GLM you could do:
> model1 anova(model1, test="F") # F test appropriate for e.g. the following values for family: "gaussian", "quasibinomial", "quasipoisson".
For gam:
> model1 anova(model1, test="F")
Regarding the inclusion of interaction terms in a GAM-model, there is e.g. the following webpage: http://stats.stackexchange.com/questions/32730/how-to-include-an-interaction-term-in-gam
Regards, G. Haege
For GLM you could do:
> model1 anova(model1, test="F") # F test appropriate for e.g. the following values for family: "gaussian", "quasibinomial", "quasipoisson".
For gam:
> model1 anova(model1, test="F")
Regarding the inclusion of interaction terms in a GAM-model, there is e.g. the following webpage: http://stats.stackexchange.com/questions/32730/how-to-include-an-interaction-term-in-gam
Regards, G. Haege
Hi there,
The anova() function will do what you want. Make sure you pick the right test when using the "test = " argument.
Best, Jorge Velez.-
Dear Jorge and Gisela,
Thanks for your guidance. I used anova() function mainly for model comparison specially in model simplification but i was not sure can i replace it with summary() function.
@ Heage are you sure we can apply "test=F" for quisipoisson and quisibinomial as well? I think "test=Chi" would fit to this families.
Warm regards,
Mehdi
According to webpage https://stat.ethz.ch/R-manual/R-devel/library/stats/html/anova.glm.html, "test=F" is appropriate for models "with dispersion estimated by moments (e.g., gaussian, quasibinomial and quasipoisson fits)". "For models with known dispersion (e.g., binomial and Poisson fits) the chi-squared test is most appropriate".
Regards, G. Haege
Dear Gisela,
Thanks for your references.do you know any references to sum up all possible "test" for GLM? Could we apply them for GLMM as well?
Warm regards, Mehdi
Hi,
I have a similar question for glmer. How to get only main factors and interactions without pairwise comparisons of their levels in glmer?
Thanks!
Janja Širovnik you can easily get them by Anova() by Car package. Please note that it must be the capital letter of 'A' in Anova(). EX:
your.model
- Badges
- Science method
- Similar topics
- Mathematics
- Statistics
- Statistical Modeling
- Analyses of Variance
- ANOVA