[R-sig-ME] SAC tests in mixed models
Vinicius Maia
v|n|c|u@@@@m@|@77 @end|ng |rom gm@||@com
Fri Jul 31 06:34:52 CEST 2020
Dear list,
As you may know, the inferences of Moran's I test on model's residuals
change when the model matrix or (X'X)^{-1} are taken into account. It
occurs because regression assumptions affect how we view the regression
residuals.
For example, Moran's I test value of spdep::lm.morantest (which uses the
model matrix) and spdep::moran.test are the same, however, their p-values
are different because the former takes into account the model matrix and
the latter doesn't.
As far as I know ncf::correlog and DHARMa::testSpatialAutocorrelation also
does not account for the model matrix. But I may be wrong about DHARMa.
In general, I think the results of moran.test, correlog and
testSpatialAutocorrelation are reliable for mixed models.
However, I have not found a test like spdep::lm.morantest for mixed models.
Do you also think the results of moran.test, correlog,
testSpatialAutocorrelation are reliable for mixed models? Does anyone know
a SAC test for mixed models which accounts for the models matrix, such as
lm.morantest?
Thanks in advance for any guidance on this topic.
Best,
Vinicius
<https://rdrr.io/cran/DHARMa/man/testSpatialAutocorrelation.html>
[[alternative HTML version deleted]]
More information about the R-sig-mixed-models
mailing list