Confidence intervals

-----BEGIN PGP SIGNED MESSAGE-----
Hash: SHA1

Dear Ben Bolker,

I am trying to calculate CI?s for the fixed effects in a lmer
model.

Relatively simple questions should *very* preferably be sent to the
r-sig-mixed-models mailing list (cc'd here), please ...

I am now using the following code

FUN <- function(fit) { return(fixef(fit)) }

result <- bootMer(Model, FUN, nsim = 5000)

bCI.1 <- boot.ci(result, index=1, type="perc") bCI.2 <-
boot.ci(result, index=1, type="perc") bCI.3 <- boot.ci(result,
index=1, type="perc") bCI.4 <- boot.ci(result, index=1,
type="perc")

Would this be a correct way?

Are you sure you want index=1 in each case?  Maybe you want
index={1,2,3,4}?

  5000 sims seems like overkill unless your model is very unstable

When I use *confint* I get the following warning:

In norm.inter(t, alpha) : extreme order statistics used as
endpoints

I don't know exactly why this warning occurs -- you would probably
need to dig in a little bit more and/or read Davison & Hinkley's book
on which the boot package is based? (based on brief googling, it seems
as though your distributions may be badly behaved/need *more*
bootstrap samples).  Plotting the bootstrap distribution, as shown in
example("bootMer"), might be a good idea.

I would highly appreciate it if you would find the time to answer
my question.

Kind Regards,

Maik Theelen

-----BEGIN PGP SIGNATURE-----
Version: GnuPG v1.4.11 (GNU/Linux)

iQEcBAEBAgAGBQJUEJ3tAAoJEOCV5YRblxUHlrIIALcA15zLu3cMMu9R7WlLo4LQ
M9CeC59ASSRdsVzW3h3LV5etsvku5XmigBMJQEi4MMfdz6avpw6ZQddRoNeHmjKb
4vJXCgO2a7sbTpz48E15ZA1YFxcDp82dxqIIrm2E2zdP4vQYXdqyAWFSuvbr4qg5
LC3DnqtSdwIR0+2XDPN6stN+Au+iWZPZ18hgfIv83xN2D7rP0GGLh9/GaSeaCXuQ
5BJLyEnoQSo9rWpySznBshSK5187orRB2dHKECL2E7AKWs4WTaPazm6QtzhDK34r
O26WN5RHDkvEDr6n3ZtWp/lAP7yo4w6FyPvASFr573S/VGiVAXpWEs64TfIt38k=
=+jy7
-----END PGP SIGNATURE-----