[R] nested model and post hoc?

[Bert Gunter]

I rather suspect the problem is primarily statistical, not R related. If at
all possible, try to get some local statistical advice. Most probably, you
have empty cells in some of the dpi.f by diss.f table.

Also, using a Gaussian family and the cells/mg response may be
inappropriate: 1 cell  in 5 mg is different than 100 cells in 500 mg. This
obviously depends on your data, which is why local help might be important.

Cheers,
Bert


On Wednesday, August 5, 2015, a_wohl <sithlord1 at gmx.net> wrote:

> Hi :-) i am really looking forward to get some help... Since I am a
> R-beginner I need it.
>
> I have a model consisting of the factors: dpi, infection, dissection day
> and
> plate. My reskponse variable is cells/mg. dpi is nested in dissection day.
> They are all fixed variables.
> I produced this nested model (best AIC index)
> glm.3<-cm.2~infect.f+dpi.f+dpi.f/diss.f, na.action=na.omit)
>
> summary(glm.3)
> Call:
> glm(formula = cm.2 ~ infect.f + dpi.f + dpi.f %in% diss.f, family =
> gaussian,
>     na.action = na.omit)
>
> Deviance Residuals:
>     Min       1Q   Median       3Q      Max
> -2.7746  -0.3625   0.0080   0.3628   1.9031
>
> Coefficients: (18 not defined because of singularities)
>                 Estimate Std. Error t value Pr(>|t|)
> (Intercept)      7.01887    0.15278  45.941  < 2e-16 ***
> infect.finf      0.05544    0.08107   0.684   0.4946
> infect.fsha     -0.15563    0.08831  -1.762   0.0791 .
> dpi.f2           0.08924    0.20459   0.436   0.6630
> dpi.f4          -0.40167    0.21082  -1.905   0.0577 .
> dpi.f8           0.01961    0.20789   0.094   0.9249
> dpi.f16          0.82210    0.20469   4.016 7.54e-05 ***
> dpi.f32          0.99639    0.21435   4.648 5.09e-06 ***
> dpi.f1:diss.f2   0.13276    0.21077   0.630   0.5293
> dpi.f2:diss.f2        NA         NA      NA       NA
> dpi.f4:diss.f2   0.46749    0.20784   2.249   0.0253 *
> dpi.f8:diss.f2  -0.06555    0.20205  -0.324   0.7459
> dpi.f16:diss.f2       NA         NA      NA       NA
> dpi.f32:diss.f2       NA         NA      NA       NA
> dpi.f1:diss.f3        NA         NA      NA       NA
> dpi.f2:diss.f3  -0.15794    0.20186  -0.782   0.4346
> dpi.f4:diss.f3   0.20921    0.21081   0.992   0.3218
> dpi.f8:diss.f3        NA         NA      NA       NA
> dpi.f16:diss.f3 -0.03611    0.20178  -0.179   0.8581
> dpi.f32:diss.f3       NA         NA      NA       NA
> dpi.f1:diss.f4   0.14040    0.20459   0.686   0.4931
> dpi.f2:diss.f4  -0.01701    0.21152  -0.080   0.9360
> dpi.f4:diss.f4        NA         NA      NA       NA
> dpi.f8:diss.f4        NA         NA      NA       NA
> dpi.f16:diss.f4 -0.19082    0.20168  -0.946   0.3449
> dpi.f32:diss.f4       NA         NA      NA       NA
> dpi.f1:diss.f5        NA         NA      NA       NA
> dpi.f2:diss.f5        NA         NA      NA       NA
>
> ........and so forth.
> Since i decided for a nested modul, I wonder where the NAs are coming from?
> Since I want to perform a post hoc test, I was trysing the glht fuinction
> form the multcomp package.
> I always get this error
> Error in modelparm.default(model, ...) :
>   dimensions of coefficients and covariance matrix don't match
>
> I think this is due to my NAs in the summary(glm.3).
> Can anyone help me, how I can perform a post hoc test (Bonferroni) on my
> data?
>
> thank you
> a_wohl
>
>
>
> --
> View this message in context:
> http://r.789695.n4.nabble.com/nested-model-and-post-hoc-tp4710784.html
> Sent from the R help mailing list archive at Nabble.com.
>
> ______________________________________________
> R-help at r-project.org <javascript:;> mailing list -- To UNSUBSCRIBE and
> more, see
> 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.
>


-- 
Bert Gunter

"Data is not information. Information is not knowledge. And knowledge is
certainly not wisdom."
   -- Clifford Stoll

	[[alternative HTML version deleted]]