Bioequivalence and Bioavailability Forum

Dear All

Anderson (2013) reported that the number of fixed effects in the model affected the results of ML and REML parameter estimation methods based on likelihood. In the study, the researcher stated that if the number of fixed effects is ≤ 4, the ML method could give better results, while if the number of fixed effects is> 4, REML would be preferred.

Based on this result and in BE studies with maximum three fixed effects ML is better

Anderson CJ (2013). Estimation: Problems & Solutions.
https://web.archive.org/web/20180827214328/http://courses.education.illinois.edu/edpsy587/lectures/estimation-beamer-online.pdf (18.01.2014).

Thank you

Mahmoud Abdel Mohsen

---

Edit: Category and subject line changed. Please don’t use series of “=” for underlining (results are different for every user); see there for the appropriate BBCode. The URL you gave is not accessible; I changed it to the archived one. Please do that yourself the next time. [Helmut]