*H*i,
I am trying to fit a GEE model with *geeglm* function. The model is the
following:
Modelo<-geeglm(sqrt ~Tra+ Mes, id=Lugar , data=datos,
family=gaussian(identity), corstr="independence")
*Tra( is a experimental treatment, 2 levels)*, *Mes* (is the month of take
data, 4 levels) and *Lugar* (is the site of study, 3 levels) are categorical
variables and *sqrt* (sqrt of Total Carbon on soil) it's a continuous
variable. I want to know if *sqrt* can be to explained for *Tra* and
*Mes*when measures among
sites (*Lugar*) are repeated measures. I get this outpout summary (A) and
anova (B).
(A) Call:geeglm(formula = sqrt~ Tra + Mes, family = gaussian(identity), data
= datos, id = Lugar, corstr ="independence")
Coefficients:
Estimate Std.err Wald Pr(>|W|)
(Intercept) 4.6733 0.7007 44.48 2.6e-11 ***
TraT1 -0.2155 0.0748 8.30 0.004 **
Mes2 -0.0159 0.0699 0.05 0.820
Mes3 -0.0596 0.1441 0.17 0.679
Mes4 -0.1581 0.0373 17.99 2.2e-05 ***
Estimated Scale Parameters: Estimate Std.err
(Intercept) 1.33 0.165
(B) Analysis of 'Wald
statistic' Table
Model: gaussian, link: identity Response: sqrt
Terms added sequentially (first to last)
Df X2 P(>|Chi|)
Tra 1 9.00e+00 0.0025 **
Mes 3 -1.37e+14 1.0000
The convectional ANOVA analysis with repeated measures give *p* values of
both between groups (cofactors: *Tra* and *Mes*) and intra groups (levels of
*Lugar*). When i try to fit the following model for know the wald and p
values of the *Lugar*'s levels I get strangers p and wald values with
summary function (A) and it’s impossible to run the anova function (B).
modelo<-geeglm(sqrt~Tra+ Lugar+ Mes, id=Lugar ,data=datos,
family=gaussian(identity), corstr ="independence")
(A)
Call:geeglm(formula = sqrt~ Tra + Lugar + Mes, family gaussian(identity), data
= datos, id = Lugar, corstr ="independence")
Coefficients:
Estimate Std.err Wald Pr(>|W|)
(Intercept) 3.23291 0.00000 Inf <2e-16 ***
TraT1 -0.23229 0.00000 Inf <2e-16 ***
Lugar2 1.97880 0.00000 Inf <2e-16 ***
Lugar3 2.31549 0.00000 Inf <2e-16 ***
Mes2 0.00147 0.00000 Inf <2e-16 ***
Mes3 -0.04217 0.00000 Inf <2e-16 ***
Mes4 -0.22612 0.00000 Inf <2e-16 ***
---
Signif. codes: 0 ‘***’ 0.001 ‘**’ 0.01 ‘*’ 0.05 ‘.’ 0.1 ‘ ’1
Estimated Scale Parameters:
Estimate Std.err
(Intercept) 0.309 0.0818
(B) anova(modelo)
Error en solve.default(vbeta[zeroidx, zeroidx, drop = FALSE]) : rutina
Lapack dgesv: sistema es exactamente singular
It’s so necessary for me get this comparisons too. Please, I appreciate if
you can help me.
Marylin Bejarano
pH student
Instituto de Ecología-UNAM
[[alternative HTML version deleted]]
