This indicates why I’d avoid bootstrapping if there’s a decent alternative…! Not least because in a highly structured experiment, you may have few replicates with identical values of all factors.
James Brown
From: GENSTAT-Request <[log in to unmask]>
On Behalf Of Murray C Hannah (DJPR)
Sent: 21 December 2020 23:34
To: [log in to unmask]
Subject: Re: Bootstrapped confidence intervals for a GLMM
And another comment … possibly to make myself unpopular, shouldn’t the bootstrap resampling reflect the random structure? i.e. Select data from F with replacement, then within F select
levels of G with replacement, … .
Has anyone seen any theory on multi-strata bootstrapping, to reassure us that this yields interpretable CIs?
Murray Hannah
Senior Research Scientist – Biometrics
| Dairy Production Sciences | Agriculture Research
Department of Jobs, Precincts and Regions
1301 Hazeldean Road Ellinbank, Victoria Australia 3821
T: (03) 5624 2260
|
M: 0408 398 234
agriculture.vic.gov.au
From: GENSTAT-Request <[log in to unmask]>
On Behalf Of Fred Newell
Sent: Monday, 21 December 2020 8:27 PM
To: [log in to unmask]
Subject: Bootstrapped confidence intervals for a GLMM
Dear All,
I’m trying to fit a generalised liner mixed model to a data set that I have using the GLMM procedure in GenStat. The random effects are in a nested design
of the type F/G and the fixed effects consist of a factorial arrangement of 2 factors A and B. I’m using a binomial distribution with a logit link function to model the response variable. My aim is to produce back-transformed means for A, B and the AB interaction
with 95% confidence intervals. While I can get the means using VPREDICT and the usual transformation, I think that I’d need to use bootstrapping to get the confidence intervals. In light of this would anyone be able to advise how to use the BOOTSTRAP procedure
in this context in order for me to be able to do this?
Regards,
Fred.
Sent from
Outlook
To unsubscribe from the GENSTAT list, click the following link:
https://www.jiscmail.ac.uk/cgi-bin/WA-JISC.exe?SUBED1=GENSTAT&A=1
To unsubscribe from the GENSTAT list, click the following link:
https://www.jiscmail.ac.uk/cgi-bin/WA-JISC.exe?SUBED1=GENSTAT&A=1