Skip to main content

Advertisement

Fig. 6 | BMC Systems Biology

Fig. 6

From: Incorporating prior knowledge improves detection of differences in bacterial growth rate

Fig. 6

The highest ranked prior depends on closeness of growth rate to the growth rate cluster. Given two clusters of four simulated curves with parameters inferred by nested sampling, we plot the evidence, logZ, for analysing a new curve using a uniform (blue), Gaussian (orange) and Cauchy (black) prior as a function of position in-between the two clusters (in terms of growth rate). In the case of the Gaussian or Cauchy priors, the solid lines indicate that cluster A has been used for prior knowledge, whilst dashed lines indicate that cluster B has been used. Data sets from cluster A had growth rates between 0.09 and 0.11 and cluster B had rates between 0.15 and 0.17. Choosing a Gaussian prior is appropriate when the target growth curve is near to the cluster, but loses evidence as curves become more dissimilar from their cluster. The Cauchy prior can alleviate this to some extent, whilst the uniform prior has little change across growth rates

Back to article page