Hi!
I am using the NLME package for R to modeling glucose-insuline
response with Bergman's model, very similar to the example in the
documentation for the NLME package.
My question concerns the model for the residuals. I use a proportional
model , Var(e_{ij})=(sigma_g*G(t))^2 for the glucose response and
Var(e_{ij})=(sigma_i * I(t))^2 for the insulin response. Hence I have
a varPower model, but I know that the power coefficient is 2, but with
different sigma coefficients for the two responses.
If I fit a model I get an estimation of the standard deviation of the
residuals, which directly can be obtained from the output: The value of the
residual 0.01966.. is very close to the real value 0.02 which I have used
for both responses in this example.
Random effects:
Formula: list(Sg ~ 1, Si ~ 1, n ~ 1, gamma ~ 1, G0 ~ 1, I0 ~ 1)
Level: id
Structure: Diagonal
Sg Si n
gamma G0 I0 Residual
StdDev: 0.5439934 0.5075674 0.3192111 0.3227774 0.08773017 0.1060001
0.01966552
However, there is only one value. Is that some kind of average over all
samples? What if I want to extract the two different sigma coefficients for
the two
responses. Just as in the example I have a level "type" that specifies
which response it is.
--
View this message in context:
http://r.789695.n4.nabble.com/NLME-error-model-with-several-responses-tp4501300p4501300.html
Sent from the R help mailing list archive at Nabble.com.
