Computing "quasi-AIC" (QAIC), in R is a minor pain, because the R Core team (or at least the ones who wrote glm, glmmPQL, etc.) are purists and don't believe that quasi-models should report a likelihood. As far as I know, there are three R packages that compute/handle QAIC: bbmle, AICcmodavg and MuMIn.The basic problem is that quasi-model fits with glm return an NA for the log-likelihood, while the dispersion parameter (ĉ, φ, whatever you want to call it) is only reported for quasi-models. Various ways to get around this are:• fit the model twice, once with a regular likelihood model (family=binomial, poisson, etc.) and once with the quasi-variant -extract the loglikelihood from the former and the dispersion parameter from the latter• only fit the regular model; extract the overdispersion parameter manually with dfun <-function(object) { with(object,sum((weights * residuals^2)[weights > 0])/df.residual) }• use the fact that quasi-fits still contain a deviance, even if they set the log-likelihood to NA. The deviance is twice the negative log-likelihood (it's offset by some constant which I haven't figured out yet, but it should still work fine for model comparisons)The whole problem is worse for MASS::glmmPQL, where (1) the authors have gone to greater efforts to make sure that the (quasi-)deviance is no longer preserved anywhere in the fitted model, and (2) they may have done it for good reason -it is not clear whether the number that would get left in the 'deviance' slot at the end of glmmPQL's alternating lme and glm fits is even meaningful to the extent that regular QAICs are. (For discussion of a similar situation, see the WARNING section of ?gamm in the mgcv package.)Example: use the values from one of the examples in ?glm:1
The purpose of this study was to investigate the nature and understanding of online talk in a New Zealand primary-school context. This research consisted of a small-scale case study of the #NZReadaloud, a pre-existing literacy programme, over 6 weeks in mid-2020. A private online group on the education platform Edmodo was established where 14 students and four teachers participating in the study could discuss a text being read aloud in their classrooms. These discussions, along with eight follow-up interviews, were analysed to seek patterns and draw conclusions about the nature of online talk in this context. A key finding from this study showed that online talk was well liked by teachers and students, but that this did not translate into greater amounts or a more interactive style of participation. Allocated time for explicit teaching and modelling of how to talk online is important. Data from this study could be used in future research or as background when planning online literacy discussions.
scite is a Brooklyn-based organization that helps researchers better discover and understand research articles through Smart Citations–citations that display the context of the citation and describe whether the article provides supporting or contrasting evidence. scite is used by students and researchers from around the world and is funded in part by the National Science Foundation and the National Institute on Drug Abuse of the National Institutes of Health.
customersupport@researchsolutions.com
10624 S. Eastern Ave., Ste. A-614
Henderson, NV 89052, USA
This site is protected by reCAPTCHA and the Google Privacy Policy and Terms of Service apply.
Copyright © 2024 scite LLC. All rights reserved.
Made with 💙 for researchers
Part of the Research Solutions Family.