Fig. 3From: Simulation-based assessment of model selection criteria during the application of benchmark dose method to quantal response dataObserved and predicted dose-response relationships for the 1-aminoanthraquinone, 2-ethylhexyl vinyl ether, and acrylamide datasets. a 1-aminoanthraquinone with a substantial weight applied for doses with high response rates (n = 13 per observed dose). b 2-ethylhexyl vinyl ether with a weight applied for doses with low response rates (n = 6 per dose). c Acrylamide with an approximately linear dose-response relationship (n = 48 per dose). Original outcomes were eosinophilic droplet in renal proximal tubular epithelium in male rats for 1-aminoanthraquinone, centrilobular hypertrophy in liver stem cells in male rats for 2-ethylhexylvinyl ether, and axon degeneration in peripheral nerve in male rats for acrylamide, which we disregard in this study. The best fit models selected using only the Akaike information criterion were the Probit model for 1-aminoanthraquinone and 2-ethylhexyl vinyl ether, and the Logistic model for acrylamide. Unbiased BMDL10 and BMD10 were estimated as 0.9 and 7.7 respectively, for 1-aminoanthraquinone, 24.7 and 28.7, respectively, for 2-ethylhexylvinyl ether, and 0.8 and 0.9, respectively, for acrylamideBack to article page