Skip to main content
Fig. 4 | BMC Bioinformatics

Fig. 4

From: Beyond the one-way ANOVA for ’omics data

Fig. 4

Simulation Study: Model misspecification. Under a B-H control of the FDR: a Percentage of responses with the correct model specification identified from each of the RFM and MRF approaches compared to the true generating process. b Percentage of responses with a model misspecification identified from each of the RFM and MRF approaches compared to the true generating process. c Percentage of responses that have been over-fitted (the fitted model includes all terms from the true generating process with additional terms) under the RFM and MRF approaches. d Percentage of responses that have been under-fitted (the fitted model includes only a subset of terms, and no others, from the true generating process) under the RFM and MRF. A model misspecification is defined to be a model that differs from the true generating process in a way not captured by over- or under-fitting. Red, green and blue correspond to simulations of 500, 1000 and 20000 responses respectively

Back to article page