lme4, failure to converge with a range of optimisers, trust the fitted model anyway?
On 5 April 2015 at 20:12, Hans Ekbrand <hans.ekbrand at gmail.com> wrote:
On Sun, Apr 05, 2015 at 07:31:25PM +1000, Ken Beath wrote:
No, you need to treat this still as binomial data, using cbind(y,n-y) as the response where y is the number of positives in each group, and n is
the
total in each group.
OK, I'll try that. What is the interpretion of the outcome in this case, is it still the logit of the probability of the outcome?
Yes.
I suggest reading one of the books that discusses fitting logistic models in R, most advanced texts have a section. Introductory Statistics with R by Peter Dalgaard has the section available in Amazon.
I actually already have that one, quite good.
You also still need a random effect for the cluster. While I'm thinking of it, should clusters and country random effects have been crossed. Generally the sampling is setup so that clusters are nested within countries which requires a different syntax.
I'm sorry but I haven't been clear on this, but the clusters are nested within countries, so there are no crossed random effects to be found.
The random effect then needs to be included as (1|country/clusterID)
*Ken Beath* Lecturer Statistics Department MACQUARIE UNIVERSITY NSW 2109, Australia Phone: +61 (0)2 9850 8516 Building E4A, room 526 http://stat.mq.edu.au/our_staff/staff_-_alphabetical/staff/beath,_ken/ CRICOS Provider No 00002J This message is intended for the addressee named and may...{{dropped:9}}