[LCMATE] informative vs non informative model comparison question

[Buczinski Sébastien]

Dear Braintrust
I'm currently collaborating on a project using LCM analysis. Basically we are trying to find regression coefficients from a logistic model where the test is measured with uncertainty (600 calves, 20% with a positive test).
We've used the approach from McInturff et al. (Modelling risk when binary outcomes are subject to error Stat Med 2004) where we asked expert to establish plausible probabilities of different association of covariate for the probability of the animal having the disease (without prior knowledge of the database).
The model is running (3 chains, 50000 iteration 5000 burnin) well and no particular problems to do it.
Then my sensitivity analysis was performed using :
1)non informative priors (beta(1,1) for my different probability profiles (vs beta distribution obtained from the experts)
2) non informative priors in both probability profiles and my imperfect test (beta (1,1)) to let the data speak.
 I used DIC as a model fit indicator.
My main model as a DIC of 567 vs 562 and 562 for the 2 non informative models.
Now what would be your advices? Should I say that I trust more my NI models than the main model? any previous papers or research dealing with this issue and how to report it?
Many thanks!


Sébastien Buczinski, Dr Vét, DÉS, MSc, DACVIM,
Professeur titulaire / Full Professor
Clinique ambulatoire bovine / Bovine ambulatory clinic
Faculté de Médecine Vétérinaire, Université de Montréal
3200 Rue Sicotte, St-Hyacinthe, J2S 2M2, Qc, Canada
Tel/Phone : 450 773 8521 (8675)
Fax 450 778 8120

" A pessimist sees the difficulty in every opportunity; an optimist sees the opportunity in every difficulty. " W. Churchill

Note: Ce courriel est destiné exclusivement au(x) destinataire(s) mentionné(s) ci-dessus et peut contenir de l'information privilégiée,confidentielle et/ou dispensée de divulgation aux termes des lois applicables. Toute utilisation, transmission ou copie non autorisée de ce courriel est interdite. Si vous avez reçu ce message par erreur, ou s'il ne vous est pas destiné, veuillez le mentionner immédiatement à l'expéditeur et effacer ce courriel.