anova.modTempEff {modTempEff} | R Documentation |
Comparing "modTempEff"
objects returned by tempeff()
using an analysis of deviance table.
## S3 method for class 'modTempEff': anova(object, ..., dispersion = NULL, test = NULL)
object,... |
fitted model objects of class "modTempEff" returned by tempeff() . |
dispersion |
currently ignored. |
test |
what sort of likelihood based criterion has to be used for model comparisons. One of
"Chisq" , "F" , "Cp" or "BIC" . |
anova.modTempEff
performs model comparisons in terms of likelihood-based criteria
depending on its argument test
. In anova.modTempEff
, test="BIC"
is also allowed.
The BIC appears to be the best choice to select the number of breakpoints. When test="Chisq"
,
likelihood ratio tests are carried out; however note that the p-values for smooth terms (included via
seas()
or csdl()
) are approximate only. The function does not work for a single
"modTempEff"
fit.
Vito Muggeo
## Not run: #continues from ?tempeff anova(o3,o1,test="Chisq") #approximate p-value.. anova(o3,o2,o1,test="Cp") ## End(Not run)