Fig. 7
From: Incorporating prior knowledge improves detection of differences in bacterial growth rate
Using prior knowledge we can attribute significance to subtle differences in growth rate. We take Escherichia coli data from ComBase and use Bayes factors to compare the curve EcBook16_22_c (black solid line and filled circles in the left panel) with a curve with subtly different growth rate, EcGB_20_b (blue solid line and filled circles in the left panel), collected at a slightly different temperature and pH. The noise level, σ, is inferred and for each curve, we use prior knowledge from a cluster of previously analysed curves (dashed lines and crosses in the left panel), measured at the same temperature and pH. Growth rates are considered to be different if there is substantial evidence for this (hypothesis 3, H 3) over equal growth rates (hypothesis 2, H 2). Due to the variability in the stochastic algorithm, we compute the Bayes factor, B 23, for 50 runs (red circles in the right panel) and compute the mean and standard deviation (black diamond and triangles in the right panel). Whilst the outcome of testing differences using an F-test is to accept the null hypotheses that the two curves have the same growth rate, a large proportion of runs for the Bayesian analysis result in the detection of a difference in growth rate