[R-sig-ME] help with glmer

[Sachiin M H]

Hi,
model1=glmer(survival~size+CON+HET+INT+(1+size|species)+(1|quadrat),offset=time,data=data,family=binomial(link="cloglog"),verbose=T)

survival- dead (1), alive(0)
size - sapling, juvenile and adult categories
HET and CON - Heterospecific and conspecific negative densitites at 20m
radius
(summ of basal area around 20 m radius of a focal tree)
INT - Census interval
quadrat - 100 number - 10x10m
time - log transformed number of years

this is the warning message at the model summary

Warning messages:1: In (function (fn, par, lower = rep.int(-Inf, n),
upper = rep.int(Inf, : failure to converge in 10000 evaluations2: In
checkConv(attr(opt, "derivs"), opt$par, ctrl = control$checkConv, :
unable to evaluate scaled gradient3: In checkConv(attr(opt, "derivs"),
opt$par, ctrl = control$checkConv, : Model failed to converge:
degenerate Hessian with 1 negative eigenvalues


my question is - how heterospecific and conspecific desnity influence
survival of sapling, juvenile and adult trees?

[[alternative HTML version deleted]]

[Steve Candy]

Hi Sachiin

I do not understand why "time" is included as an offset and even before
that what the variable "number of years" refers to. An offset does not
allow any scaling by a corresponding regression coefficient so you have to
be very careful to determine whether a predictor variable can validly be
included as an offset.

cheers

Steve
--
Dr Steven G. Candy
Director/Consultant
SCANDY STATISTICAL MODELLING PTY LTD
(ABN: 83 601 268 419)
70 Burwood Drive
Blackmans Bay, TASMANIA, Australia 7052
Mobile: [phone]

[[alternative HTML version deleted]]

[Ben Bolker]

This post might be inappropriate. Click to display it.

[Ben Bolker]

An off-list conversation with Steve convinces me that he is right and
I'm wrong (on several counts); I was reading too fast.

* 'time' is indeed log-transformed. BUT ...
* for the standard exposure correction, it would be log(INT<="census
exposure">) and not log(time<="number of years">) that would be the
sensible thing to do.

Post by Steve Candy
Hi Sachiin
I do not understand why "time" is included as an offset and even before
that what the variable "number of years" refers to. An offset does not
allow any scaling by a corresponding regression coefficient so you have to
be very careful to determine whether a predictor variable can validly be
included as an offset.
cheers
Steve