[R-sig-ME] GLS with small-sample-sizes corrections

[j@z@vr@kidis m@ili@g off @ki@@l]

Dear all,

My question is more theoretical I guess and less practical. Is it possible/necessary/appropriate to apply small-sample-size corrections, such as Satterthwaite and Kenward-Roger approximation, in a marginal model via GLS ? So no random parts, but only covariance structure... I am working with datasets of 5-10 observations per group. I already use the REML instead of the conventional ML, but I was wondering if this is not enough. Although a quick check I did, gave small to none differences between with and without these corrections... 

And of course, if the answer is "yes, you should apply them", can you tell me if and how can I do it in R with lme4/nlme ? I currently use the nlme::gls() function...


Kind regards,

John Zavrakidis

Junior Researcher - Statistician
Department of Epidemiology and Biostatistics

e-mail:  j.zavrakidis using nki.nl





The Netherlands Cancer Institute | Plesmanlaan 121 | 1066 CX AMSTERDAM | www.nki.nl



This e-mail is intended for the addressee(s) eyes only. If you are not the intended recipient, you are hereby kindly requested to inform the sender of this. In view of the electronic nature of this communication, The Netherlands Cancer Institute (NKI) is neither liable for the proper and complete transmission of the information contained therein nor for any delay in its receipt. For information about the Netherlands Cancer Institute, go to www.nki.nl.


Dit e-mailbericht is uitsluitend bestemd voor de geadresseerde(n). Als dit bericht niet voor u bestemd is, wordt u vriendelijk verzocht dit aan de afzender te melden. Het Antoni van Leeuwenhoek  (AVL) staat door de elektronische verzending van dit bericht niet in voor de juiste en volledige overbrenging van de inhoud, noch voor tijdige ontvangst daarvan. Voor informatie over het AVL raadpleegt u www.avl.nl