[R] Advice in model construction

[Weidong Gu]

Hi Chris,

Linear regression model of categorical variables is equivalent to
anova. If one of estimates of coefficients is significant in lm, it is
interpreted as samples of Family did not come from the same population
and the signficant factor level is compared to the reference level. If
you want pair-wise comparison, you need post-hoc testing methods
(?TukeyHSD).

HTH

Weidong Gu

On Wed, Oct 5, 2011 at 1:55 PM, Chris Mcowen <chrismcowen at gmail.com> wrote:
> Dear list,
>
> I am unsure how to structure my model, i have tried something and it makes sense but i am unsure if i am interpreting it correctly?
>
> i have a continuous response variable - the observed quantity of evolutionary history - EH
>
> Then i have a number of species which have a hierarchical structure ~ Genus, Family etc
>
> My research question is do certain families have significantly higher ( or lower) EH values than the others.
>
> Reproducible example:
>
> example <- structure(list(Family = structure(c(2L, 1L, 1L, 5L, 7L, 7L, 3L,
> 4L, 6L, 6L, 1L, 3L), .Label = c("Araceae", "Asphodelaceae", "Bromeliaceae",
> "Cyperaceae", "Orchidaceae", "Poaceae", "Zingiberaceae"), class = "factor"),
>    Genus = structure(c(3L, 4L, 4L, 1L, 6L, 6L, 2L, 5L, 8L, 9L,
>    4L, 7L), .Label = c("Acianthera", "Aechmea", "Aloe", "Anthurium",
>    "Bulbostylis", "Hedychium", "Lindmania", "Psathyrostachys",
>    "Sesleria"), class = "factor"), Species = structure(c(9L,
>    1L, 10L, 11L, 7L, 4L, 3L, 5L, 8L, 2L, 6L, 12L), .Label = c("bonplandii",
>    "coerulans", "cymosopaniculata", "elatum", "emmerichiae",
>    "gehrigeri", "glabrum", "juncea", "pubescens", "sagittatum",
>    "scalpricaulis", "sessilis"), class = "factor"), EH = c(8.746525,
>    24.462699, 33.03942, 32.719489, 13.598201, 13.598201, 13.164928,
>    9.339228, 9.69705, 13.478372, 37.497137, 59.562911)), .Names = c("Family",
> "Genus", "Species", "EH"), class = "data.frame", row.names = c(NA,
> -12L))
>
> #My model
>
> test <- lm(EH~Family, data = example)
>
> #in this small example no families are significant but if one was - would that mean they have significantly more EH than the others?
>
> Thanks
>
> Chris
>
>
>
> ______________________________________________
> R-help at r-project.org mailing list
> https://stat.ethz.ch/mailman/listinfo/r-help
> PLEASE do read the posting guide http://www.R-project.org/posting-guide.html
> and provide commented, minimal, self-contained, reproducible code.
>