Simulating points from GLM corresponding to new x-values
Hi Cliff -- thanks for the suggestion. I tried extracting the predicted mean and standard error using predict(). Afterwards I simulated the dependent variable using rnorm(), with mean and standard deviation taken from the predict() function (sd = sqrt(n)*se). The points obtained this way were scattered far too much (compared to points obtained with simulate()) -- I am not quite sure why. Unfortunately the documentation of the simulate() function does not provide much information about how it is implemented, which makes it difficult to judge which method is best (predict() or simulate(), and it is also unclear whether simulate() can be applied to glms (with family=gaussian or binomial). Any suggestions for how to proceed? Jacob
On 12 Aug 2009, at 13:11, Clifford Long wrote:
Would the "predict" routine (using 'newdata') do what you need? Cliff Long Hollister Incorporated On Wed, Aug 12, 2009 at 4:33 AM, Jacob Nabe- Nielsen<jacobnabe at me.com> wrote:
Dear List,
Does anyone know how to simulate data from a GLM object correponding
to values of the independent (x) variable that do not occur in the
original dataset?
I have tried using simulate(), but it generates a new value of the
dependent variable corresponding to each of the original x-values,
which is not what I need. Ideally I whould like to simulate new
values
for GLM objects both with family="gaussian" and with
family="binomial".
Thanks in advance,
Jacob
Jacob Nabe-Nielsen, PhD, MSc
Scientist
--------------------------------------------------
Section for Climate Effects and System Modelling
Department of Arctic Environment
National Enviornmental Research Institute
Aarhus University
Frederiksborgvej 399, Postbox 358
4000 Roskilde, Denmark
email: nabe at dmu.dk
fax: +45 4630 1914
phone: +45 4630 1944
[[alternative HTML version deleted]]
______________________________________________ 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.